CEOP-AEGIS Periodic Report #1

CEOP-AEGIS (GA n° 212921) Periodic Report no. 1

PROJECT PERIODIC REPORT

Grant Agreement number: 212921

Project acronym: CEOP-AEGIS

Project title: Coordinated Asia-European long-term Observing system of Qinghai – Tibet
Plateau hydro-meteorological processes and the Asian-monsoon systEm with Ground satellite
Image data and numerical Simulations

Funding Scheme: CP-SICA

Date of latest version of Annex I against which the assessment will be made: 25/08/2009

Periodic report: 1st x 2nd ! 3rd ! 4th !
Period covered: from 1/5/2008 to 31/10/2009

Name, title and organisation of the scientific representative of the project's coordinator1:

Prof.dr. Massimo Menenti Faculty of Aerospace Engineering, TU Delft, Delft, The Netherlands

Tel: +31 15 2784244

Fax: +31 15 278348

E-mail: M.Menenti@tudelft.nl

Project website2 address: http://www.ceop-aegis.org/

1
Usually the contact person of the coordinator as specified in Art. 8.1. of the grant agreement
2
The home page of the website should contain the generic European flag and the FP7 logo which are available in electronic
format at the Europa website (logo of the European flag: http://europa.eu/abc/symbols/emblem/index_en.htm ; logo of the 7th
FP: http://ec.europa.eu/research/fp7/index_en.cfm?pg=logos). The area of activity of the project should also be mentioned.

Page 1 of 98


Declaration by the scientific representative of the project coordinator1

1
I, as scientific representative of the coordinator of this project and in line with the obligations as stated in
Article II.2.3 of the Grant Agreement declare that:

! The attached periodic report represents an accurate description of the work carried out in this project for
this reporting period;

! The project (tick as appropriate):

x has fully achieved its objectives and technical goals for the period;

! has achieved most of its objectives and technical goals for the period with relatively minor
3
deviations ;

! has failed to achieve critical objectives and/or is not at all on schedule . 4

! The public website is up to date, if applicable.

! To my best knowledge, the financial statements which are being submitted as part of this report are in line
with the actual work carried out and are consistent with the report on the resources used for the project
(section 6) and if applicable with the certificate on financial statement.

! All beneficiaries, in particular non-profit public bodies, secondary and higher education establishments,
research organisations and SMEs, have declared to have verified their legal status. Any changes have
been reported under section 5 (Project Management) in accordance with Article II.3.f of the Grant
Agreement.

Name of scientific representative of the Coordinator1: .Prof. Dr. Massimo Menenti.

Date: ....21..../ ...12......./ 2009......

Signature of scientific representative of the Coordinator1:

3
If either of these boxes is ticked, the report should reflect these and any remedial actions taken.
4
If either of these boxes is ticked, the report should reflect these and any remedial actions taken.

Page 2 of 98


Project Title

Coordinated Asia-European long-term Observing system of Qinghai – Tibet
Plateau hydro-meteorological processes and the Asian-monsoon systEm with
Ground satellite Image data and numerical Simulations
CEOP AEGIS

Thematic Priority: ENV.2007.4.1.4.2. Improving observing systems for water resource management

Start Date of the Project: 1 – May – 2008 Duration: 48 months

Report Title
1st Periodic Report
May 1st 2008 – October 31st 2009

Massimo Menenti1, Li Jia2 and Jerome Colin3
1 Faculty of Aerospace Engineering, TU Delft, Delft, The Netherlands,
2 Alterra, Wageningen University and Research Centre, Wageningen, The Netherlands
3 Image Sciences, Computing Sciences and Remote Sensing Laboratory, University of Strasbourg, Illkirch,
France

Date: December 21st 2009
Version: 1.0

Page 3 of 98


Coordinator contact details

Prof.dr. Massimo Menenti

E-mail: M.Menenti@tudelft.nl
Web site: www.lr.tudelft.nl/olrs
Telephone: +31 15 2784244 Fax: +31 15 278348

Deputy coordinator details:
Dr. Li Jia
E-mail: li.jia@wur.nl
Web site: http://www.alterra.wur.nl/UK/
Telephone: +31 317 481610 Fax: +31 317 419000

Contractors involved

BENEFICIARY BENEFICIARY NAME BENEFICIARY COUNTRY DATE DATE EXIT
NUMBER SHORT NAME ENTER PROJECT
PROJECT
1 CO Université de Strasbour LSIIT UDS France 1 48
2 CR International Institute for Geo- ITC The 1 48
information science and Earth Netherlands
Observation
3 CR ARIES Space ARIES Italy 1 48
4 CR University of Bayreuth UBT Germany 1 48
5 CR Alterra - Wageningen University ALTERRA The 1 48
and Research Centre Netherlands
6 CR University of Valencia UVEG Spain 1 48
7 CR Institute for Tibetan Plateau ITP China 1 48
Research – Lhasa, Tibet
8 CR China Meteorological CAMS China 1 48
Administration – Beijing
9 CR Beijing Normal University– BNU China 1 48
Beijing
11CR University of Tsukuba – UNITSUK Japan 1 48
12 CR WaterWatch WAWATCH The 1 48
Netherlands
13 CR Cold and Arid Regions CAREERI China 1 48
Environmental and Engineering
Research Institute – Lanzhou,
Gansu
14 CR University of Ferrara UNIFE Italy 1 48
15 CR Institute of Geographical IGSNRR China 1 48
Sciences and Natural Resources
Research CAS – Beijing
16 CR Institute for Remote Sensing IRSA China 1 48
Applications CAS – Beijing
17 CR Future Water FUWATER The 1 48
Netherlands
18 CR Delft University of Technology TUD The 12 48
Netherlands
19 CR National Institute of Technology NIT India 12 48

Page 4 of 98


CO = Coordinator CR = Contractor

1. Publishable summary
CEOP AEGIS 1st Periodic Report: May 1st 2008 – October 31st 2009
Summary

http://www.ceop-aegis.org/

Objectives
The goal of this project is to:
1. Construct out of existing ground measurements and current / future satellites an observing
system to determine and monitor the water yield of the Plateau, i.e. how much water is finally going
into the seven major rivers of SE Asia; this requires estimating snowfall, rainfall, evapotranspiration
and changes in soil moisture;
2. Monitor the evolution of snow, vegetation cover, surface wetness and surface fluxes and
analyze the linkage with convective activity, (extreme) precipitation events and the Asian Monsoon;
this aims at using monitoring of snow, vegetation and surface fluxes as a precursor of intense
precipitation towards improving forecasts of (extreme) precipitations in SE Asia.

Work Performed
The project started with a kick-off meeting held in Beijing on May 1st – 5th 2008 attended by 65
participants. In preparation of the meeting all partners were requested to define more precisely their
contribution and roles. This material provided a good basis for a productive meeting. A project mailing
list system was established to handle internal communication, given the complexity of the consortium.

The 1st Annual Progress Meeting was held in Milano, Italy on June 29th through July 3rd, including a
joint workshop with the CEOP High Elevation Initiative (HE) and an internal businness meeting
dedicated to a review of progress and to the preparation of the 1st Periodic Report. The meeting was
attended by 30 participants. In preparation of the meeting all partners were requested to prepare an
overview presentation for each Work Package.

The material prepared for the meetings is available on the project web site. To date there are 112
registered Team Members.

During the 1st six months period work focused on three main objectives:
1. Define the work plan and detailed contributions of partners;
2. Perform local experiments and collect first data for validation of algorithms and models;
3. Review and improvements of algorithms and models.
Ad.1. In order to identify more precisely roles and responsibilities all partners were requested to
elaborate further the work plan now included in the Description of Work. This includes now a more
precise description of (sub)-tasks and of elements of contractual deliverables with individual
responsibilities clearly identified.
Ad.2. Field experiments were carried out during the reporting period as described under “Main
Results” below
Ad.3 Work towards improvement of retrieval algorithms, process models and land-atmospheric models
advanced in several directions. This included collection and preparation of data sets acquired by space-
and airborne platforms to test algorithms and models, numerical experiments to document the
performance of algorithms and process models and improvement of algorithms and models in those
cases where the causes of poor performances was known already. More details are provided under

Page 5 of 98


“Main Results” below.

During the 2nd six-months period work focused on five main objectives:
1. Finalize and implement Grant Agreement, including accession of partners;
3. Review and improvements of algorithms and models.
4. Design, development and use of atmospheric and water balance models;
5. First analyses of time series of drought and flood indicators
Ad.1. The Grant Agreement was completed and signed on December 4th 2008. Accession forms were
signed by all partners except Partner NIH. As explained below, the National Institute of Technology,
Rourkela, will replace NIH and carry out all planned tasks.
Ad.2. Field experiments were carried out during the reporting period and data analysis started as
described under “Main Results” below. Work concentrated on the analysis of ground measurements on
land – atmosphere interactions collected at the permanent observatories on the Plateau.
Ad.3 Work towards improvement of retrieval algorithms, process models and land-atmospheric models
advanced towards the implementation of specific improvements emerged in the previous period. This
included development of new procedures to deal with complex terrain in radiative transfer models and
retrieval algorithms, new algorithms for the retrieval of land surface temperature and radiative fluxes at
the surface and preparation of data sets on precipitation measurements with rain radars.
Ad.4 Work advanced both on the assessment of connections between land surface conditions with
convective activity and precipitation events and on the design of the regional water balance model to
integrate all observations for the entire Plateau.
Ad. 5 Work was also initiated on the analysis of time series of satellite data towards the early detection
of anomalies in land surface conditions and early warning on droughts and floods. Because of the need
for extended data records, this element of the project relies on existing data sets, besides the ones
generated by the project. During this 6-months period work concentrated on development of
procedures for the detection of anomalies, based on a moving window analysis and comparison with
the climatology of the land surface variables under consideration. Different indicators were evaluated.

During the 3rd six-months period work focused on the same five main objectives as in the 2nd six-
months period:
1. Finalize documents for the amendments of the Grant Agreement, including accession of new
partners;
3. Improvements of algorithms and models.
4. Development and use of atmospheric and water balance models;
5. First analyses of time series of drought and flood indicators.
Ad.1.The access of two new partners, i.e. NIT and TUD required a significant amount of time and
work. Progress of the project was monitored through a series of Skype conferences, the 1st Annual
Progress Meeting and additional working meetings in 2009: Beijing August and October, Lanzhou in
August and Roorkee in September.
Ad.2. Field work intensified during this period. In addition to the normal operation of the
observatories, new instruments were installed to improve observations of radiative and turbulent heat
fluxes and to characterize the size distribution of rain droplets, necessary to improve accuracy of
retrievals by rain radars (see Main Results below).
Ad.3 Work towards improvement of retrieval algorithms was focused on atmospheric correction of
satellite measurements in the VNIR-SWIR, TIR and microwave regions. This included dealing with
retrieval of Land Surface Variables using data acquired by the new satellites HJ-1B (China) and IRS
(India). The new algorithms developed in the previous period were applied to generate time series of
Snow Covered Area and Snow Water Equivalent. The development of a new data processing system
for Surface Energy Balance analyses based on the combination of satellite measurements with PBL

Page 6 of 98


fields generated by the GRAPES NWF model was completed.
Ad.4 Significant advances have been achieved towards analysis of land – atmosphere interactions with
atmospheric models and towards regional modeling of the Plateau water balance. A full forecast run
was performed for the entire year 2008 with the system GRAPES. A study of the sensitivity Monsoon
Convective Systems (MCS) to land surface conditions was carried out with the model WRF at the
University of Tsukuba and the prototype water balance model of the Qinghai –Tibet Plateau was
implemented at a 5 km x 5 km spatial resolution and applied to obtain daily rainfall excess and river
flow over the entire domain
Ad.5 Several results became available on different indicators relevant to drought and flood early
warning. Work focused on two parallel streams: improving algorithms and analyzing available time
series of satellite data. A new version of the HANTS algorithm was released and a new model to
compute daily EvapoTranspiration (ET) was developed and applied. Time series of satellite data on
Land Surface Temperature (LST), photosynthetic activity (EVI, fAPAR) and soil wetness were
analyzed to document inter-annual variability, detect anomalies and evaluate them as precursor
indicators for drought and flood early warning.

Main Results
Field experiments During the 1st Reporting Period the existing system of Plateau observatories was
improved by adding several instruments: gauges to measure total precipitation above 6000 m, two
Long Path Scintillometers, three disidrometers to measure the size distribution of water droplets, four
sets of radiometers to measure the four components of the radiative balance and one suntracker to
measure direct irradiance.
Several Co-Investigators participated in a major RS experiment covering an entire watershed on the
northern rime of the Plateau: the WATER project provided invaluable detailed data to improve and
validate several algorithms to be used within CEOP-AEGIS. Collection of soil moisture and
temperature measurements at the Maqu site for the validation of algorithms to retrieve soil moisture
continued. An expedition to the the Yamdruk-tso lake basin and Qiangyong Glacier was carried out.
The Naimona'Nyi ice core was processed in cold room.
The first eddy-covariance measurements of turbulent flux densities became available after quality
characterization and gap filling. The analysis of the data collected at the NamCo observatory revealed
a significantly higher number of free convection events in the monsoon period. The results have been
published in JGR. An approach to upscale flux measurements to the grid scale of meso-scale models
and remote sensing data was developed.

Work towards improvement of retrieval algorithms, process models and land-atmospheric models
advanced in several directions:
- Collection and preparation of several data sets comprising multi-spectral, multi-angular
radiometric data;
- Evaluation of land – atmosphere models
- Review and preparation of codes of radiative transfer models of the soil-vegetation system
- Improvement and generalization of multi-scale model of land surface energy balance;
- Estimation and mapping of land – atmosphere heat and water exchanges with ASTER multi-
spectral radiometric data for the areas surrounding the ITP observatories on the Plateau;
- Preparation of microwave radiometer data (AMSR-E) for the evaluation of soil moisture
retrieval algorithms;
- Improvement of model to characterize the diurnal cycle of Land Surface Temperature using
Feng Yun infrared data and use of the CLM to relate the diurnal LST cycle to soil moisture
- Improvements in the meso-scale land-atmospheric model GRAPES of CMA; preliminary case
studies performed and hypotheses identified;
- Preparation of data sets for the evaluation of candidate water balance models; evaluation of
snow-melt-runoff models using MODIS and AMSR-E satellite data;

Page 7 of 98


- Preparation of MODIS time series (LAI/fAPAR, Vis, and LST) for entire China;
- Improvements in the algorithms to detect and predict anomalies in vegetation development;
- Case studies on drought events combining ground and satellite data;

2nd period
- Analysis of sample data set HeiHe basin with simultaneous multi-angular, multi-spectral and
lidar observations of vegetation canopies
- Topography correction inserted in the RT modeling system for the VNIR, SWIR and TIR
spectral ranges.
- Development of a simple model to describe the thermal directional radiation in rugged terrain;
- A topographic correction algorithm for albedo retrieval in rugged terrain was developed.
- Development of a preliminary algorithm to calculate land surface temperature from AMSR-E
data;
- Developing the concept of a new radiative transfer model, capable of simulating the seasonal
changes of canopy structure;
- Development of new version of MSSEBS (vers. 2.0.2) SEB algorithm;
- Development of algorithm for regional estimation of net radiation flux;
- Determination the surface albedo, surface temperature, vegetation fractional cover, NDVI,
LAI and MSAVI over whole Tibetan Plateau;
- Implementation of a radiative transfer soil moisture retrieval method using ASCAT data
- Comparison of in situ data collected by Maqu soil moisture monitoring network with AMSR-E
VUA-NASA satellite soil moisture products;
- Collected the soil moisture and temperature data of 20 SMTMS, and replaced 4 temperature
and moisture probes;
- Processing the raw precipitation radar data in the Tibetan Plateau and provide the gridded
precipitation data for case studies;
- Final revision of paper on the nighttime monsoon precipitation over the TP was submitted to
JMSJ and accepted in March
- Simulation of daily snow cover using daily and eight-day MODIS snow cover products and
meteorological observation;
- Analysis of glacier and lake changes using observed data and RS data in the Nam Co Basin;.

3rd period
Development of algorithms and retrieval of canopy structure from airborne LIDAR;
- Development of algorithm for atmospheric corrections of AMSR-E (microwave);
- Generalized procedure for atmospheric correction based on an ensemble of MODTRAN
simulations;
- Automation of procedures to generate LST from MODIS data;
- Implementation and first tests on generic algorithm for retrieval of LAI and fCover;
- Development of new algorithm to retrieve LST from HJ-1B (China) and IRS (India) data;
- Development of new Angular & Spectral Kernel based BRDF model for the normalization of
data acquired with different angular and spectral configurations;
- First test of SEB algorithm combining satellite data for land surface observations and PBL
fields generated with high resolution atmospheric model (GRAPES);
- Evaluation against turbulent heat flux measurements of SEB estimates based on ASTER data;
- Mosaic of rain-rates observed with rain radars over the Plateau have been generated and
delivered to other investigators for calibration of algorithms based on satellite data;
- Improved algorithm for retrieval of snow covered area from MODIS
has been developed and evaluated against observations at higher spatial resolution ( TM);

Design, development and use of atmospheric and water balance models.

Page 8 of 98


2nd period
- The first numerical experiments with the GRAPES land – atmosphere modeling and data assimilation
system were performed and evaluated:
- Sensitivity experiments of different soil initial conditions on the development of convections
by using 2-km resolution of GRAPES_Meso
- Detection of Meso-scale Convective System (MCS) on the TP was done for the passed six
years using METEOSAT-IR data
- Preparing GIS files for hydrological modeling, including boundary, DEM. Slope, aspect,
stream network.
-Model selection and algorithm comparison report for Plateau water balance monitoring tool was
completed

3rd period.
- The system GRAPES of CMA has been applied to generate forecasts for the entire year 2008;
- A study on the sensitivity of MCS to land surface heating has started using the WRF numerical
model at the Univ. Tsukuba;
- Gridded climate data have been used to compute the water balance of the Headwaters of the
Yellow River Basin and to compute potential ET;
- The prototype of the Qinghai – Tibet Plateau distributed water balance model has been
implemented and applied to compute for the year 2000 daily water balance for each 5 km x 5 km grid
and water routing; model riverflow at seven selected sections is being compared with observations;
-Model parameterization of glaciers mass balance is being applied to the Zhadang glacier; in- depth
case – study including the use of satellite data is in progress;

Analyses of time series of drought and flood indicators
2nd period
-Available satellite data were retrieved, time series were constructed and first analyses were
performed:
-Algorithm development on drought monitoring by time series analysis of anomalies in several land
surface parameters;
-Using time series of VTCI AVI, VCI and TCI as indicators for the estimation of the drought
impacts;
-Analysis of time series meteorological data (air temperature and precipitation, wind speed, air
humidity, solar radiation, etc)
-Development of soft computing techniques based on ANN and Fuzzy logic model for real time flood
forecasting

3rd period
- A new version of the HANTS algorithm for time series analysis of satellite data has been
released;
- A multi-annual MODIS data set covering the Plateau and surrounding regions has been created
after improved cloud screening and used to compute at-surface net radiation in addition to LST, EVI
and fAPAR;
- Analysis of a 25 years climatology of AVHRR LST and NDVI has been completed;
- Time series of SPI and VTCI have been generated and used as an indicator for drought
forecasting;
- A new ET model has been applied to evaluate potential yield loss in the winter 2008;
- A first evaluation of AMSR-E time series as an indicator of soil wetness and to detect
(positive) anomalies has been completed for the Plateau and Northern India;

Expected Final Results

Page 9 of 98


Data base containing ground observations, satellite data and higher level products, hydrologic
and atmospheric model fields for the period 2008 – 2010 over the Qinghai – Tibet Plateau.
System to generate daily streamflow in the upper catchment of all major river in SE Asia
gridded to 5 km x 5 km.
Potential Impact and Use of Results
Implementation and demonstration of an observing system of water balance and water flow on
and around the Qinghai – Tibet Plateau will provide to all countries information on water
resources and the role of the Plateau in determining weather and climate in the region.

2. Project objectives for the period
The goal of this project is to:
1. Construct out of existing ground measurements and current / future satellites an observing system to
determine and monitor the water yield of the Plateau, i.e. how much water is finally going into the seven major
rivers of SE Asia; this requires estimating snowfall, rainfall, evapotranspiration and changes in soil moisture;
2. Monitor the evolution of snow, vegetation cover, surface wetness and surface fluxes and analyze the
linkage with convective activity, (extreme) precipitation events and the Asian Monsoon; this aims at using
monitoring of snow, vegetation and surface fluxes as a precursor of intense precipitation towards improving
forecasts of (extreme) precipitations in SE Asia.

During the first year of the project, emphasis in all WP-s will be on review tools, experimental protocols,
algorithms and models. On this basis, the elements of the investigations next step will be identified in detail: the
first detailed description of new retrieval algorithms will be available, data analysis protocols will be agreed,
modelling experiments will be designed and the organization of data base will be consolidated.

During the second year of the project, work will be focused on the Algorithms Theoretical Basis Documents and
potential progresses towards community model to determine land-atmosphere energy and water fluxes with
multi-spectral satellite images. First analysis of datasets with candidate algorithms and models will be presented,
with preliminary results on time series analysis of Plateau water balance, droughts and floods indicators.

Page 10 of 98


3. Work progress and achievements during the period
Please provide a concise overview of the progress of the work in line with the structure of Annex I of the Grant Agreement.

For each work package -- except project management, which will be reported in section 3.5--please provide the
following information:

• A summary of progress towards objectives and details for each task;

• Highlight clearly significant results;

• If applicable, explain the reasons for deviations from Annex I and their impact on other tasks as well as on
available resources and planning;

• If applicable, explain the reasons for failing to achieve critical objectives and/or not being on schedule and explain
the impact on other tasks as well as on available resources and planning (the explanations should be coherent with
the declaration by the project coordinator) ;

• a statement on the use of resources, in particular highlighting and explaining deviations between actual and
planned person-months per work package and per beneficiary in Annex 1 (Description of Work)

• If applicable, propose corrective actions.

Page 11 of 98


3.1 Work progress in WP 1 and achievements during the period
" A summary of progress towards objectives and details for each task
Task 1.1
The in-situ data has been collected in the observation network of the GAME/Tibet and CAMP/Tibet and
the Mt. Everest station(QOMS), the Nam Co station(NAMOR) and the Linzhi Station(SETS) of the
TORP(Tibetan Observation and Research Platform) and Namco site of Tip( formally KEMA Station of TiP).
Four components radiation system were set up at the sites of D110, MS3608, Namco area, and Lhasa branch
of ITP (formally Yakou of Namco). Field trip to the Yamdruk-tso lake basin and Qiangyong Glacier was
performed. Precipitation, lake water and river water samples has been collected at 3 stations in this basin for
isotope analysis in the laboratory in Beijing. Glacier shallow ice cores were drilled at 6100m of the glacier to
rebuild the annual precipitation data in high elevation region. Daily atmospheric vapor samples were collected
at Lhasa and are still on going. Fig.1 to Fig.4 are the sites layout and the stations of this WP.

(a) (b)
Fig.1.1 The geographic map and the sites layout during the GAME/Tibet and the CAMP/Tibet.
(a) GAME/Tibet; (b) CAMP/Tibet.

Fig.1.2 The instruments in Mt.Everest station, Namco station and Linzhi station of ITP/CAS

Page 12 of 98


Fig.1.3. Sites of the four components radiation system over the Tibetan Plateau.

The seasonal and inter-annual time scale of the exchange of surface heat flux, momentum flux, water
vapour flux, surface and soil moisture over the different land surfaces of the Tibetan Plateau, and the structure
characteristics of the Surface Layer (SL) and Atmospheric Boundary Layer (ABL) were analyzed in the last one
and half year. The aerodynamic and thermodynamic variables were determined over the different land surfaces
of the Tibetan Plateau. The characteristics of precipitation and atmospheric water vapour transport over and
surrounding the Tibetan Plateau area were analyzed.
Task 1.2:
A technical report was prepared for the documentation of the flux calculation procedure in order to provide
all users of flux data the necessary information. Furthermore, within an UBT field trip to the Tibetan Plateau
(June-August 2009) a workshop was held from June 29th to July 1st for participants of ITP and CAREERI about
the usage of the UBT software packages for EC data post processing, footprint and QA/QC techniques. This
ensures a uniform data processing for all ground truth EC stations related to CEOP AEGIS, which is the task of
ITP and CAREERI according to the data policy rules.
Task 1.3:
In order to apply detailed footprint analysis for the EC stations, all necessary site information to prepare the
required land use maps were collected for Bj, Namco and Qomolangma site during the UBT field trip in summer
2009. Detailed footprint analysis already exists for Namco in late 2005 and from Oct 2005 up to Sept 2006, but

Page 13 of 98


has to be refined with actual flux data. Missing site information for Linzhi station will be gathered during a post
workshop excursion in July 2010 right after the CEOP workshop in Lhasa and the calculation of the footprint
analysis starts as soon as the flux data is available.

Task 1.4:
The gap filling will be processed following a procedure developed by Ruppert et al., but an extension to the
latent heat flux has to be made, for which data from Tibetan Plateau are necessary. The procedure starts as soon
as the flux data is available.

Task 1.5:
In order to find an adequate path for LAS measurements at Qomolangma site possible solutions were
investigated during the UBT field survey in summer 2009. The LAS system was set up in Mt.Everest
(Mt.Qomolangma) station in November, 2009 (Fig.4).Afterwards a preliminary footprint report was elaborated
examining the possible paths and hinting at the optimal solution. The results were documented within a special
report, the selected path and its respective footprint is shown in Fig.5.

Fig.4 The LAS system in Mt.Qomolangma(Mt.Everest ) Station

Fig.5: Selected path (solid red line) for the LAS measurements at Qomolangma site with source contributions for a
footprint “climatology” of the expected wind distribution, unstable stratification, zm = 20m.

A set of LAS was installed and aligned in Naqu BJ station (31°22'7.18"N, E91°53'55.36"E) in July, 2009, Naqu
area of Tibet. The underlying surface of observation site is alpine meadow. The effective height and path length
is 8.63 m and 1560m, respectively. Combined with the measurements of Eddy Covariance system (EC) and
Automatic Weather Station (AWS), the performance of LAS under Tibetan plateau environment has been
checked.

Page 14 of 98


Fig.6 The LAS system in Naqu BJ station

Task 1.6:
A first error analysis of flux data was given in a technical report. This will be updated as soon as the flux data is
available.

Task 1.7:
For tasks 1.6 and 1.7 a footprint scheme is currently developed by UBT and will soon be published in a peer
reviewed journal. A foundation for this scheme was elaborated within a Master thesis, for a description see
section results. Furthermore, a experiment was performed nearby the Namco Station (Fig.7). The investigations
cover EC, energy balance and soil moisture measurements for a period from June 26th to August 8th and was set
up directly at the shoreline of a small lake, pre-located to the Namco lake. This measurements will be used to
validate the footprint related upscaling scheme and serve for parameterization of fluxes above lake surface and
Kobresia mats. A documentation of the experiment is now available.

Fig.7. Turbulence measurements at Namco lake

Page 15 of 98


• Highlight clearly significant results

1. Underlying surface roughness lengths under the quality control of observation were determined

Eddy covariance flux data collected from ITP/CAS three research stations (Qomolangma station, Namco
station and Southeast Tibet station-Linzhi station) on the Tibetan Plateau are used to analyze the variation of
momentum transfer coefficient (CD), heat transfer coefficient (CH), aerodynamic roughness length (z0m), thermal
roughness length (z0h) and excess resistance to heat transfer (kB-1). All the data was checked under the
quality control firstly. The monthly average surface roughness, bulk transfer coefficient and excess resistance to
heat transfer at all three sites are obtained. Momentum transfer coefficient (CD) is quite changeable during the
day but relatively stable and lower in the night. The parameter kB-1 exhibits clear diurnal variations with lower
values in the night and higher values in the daytime, especially in the afternoon. Negative values of kB-1 are often
observed in the night for relatively smooth surfaces on the Tibetan Plateau.

Page 16 of 98


(a)

(b)

Fig. 8 Frequency distribution of ln(z0m) at Nam Co station in September(a) and October(b)

Page 17 of 98


(a) (b)

(c)

Fig. 9 The diurnal variations of observed excess resistance to heat transfer (kB-1)at Qomolangma station(a), Namco station(b)
and Southeast Tibet station(c) in March

2 Variation characteristics of radiation of the wetland surface in the Northern Tibetan Plateau
Based on the observed data at Automatic Weather Site(AWS) of MS3478 in the typical wetland of northern
Tibetan Plateau from March 2007 to February 2008. The seasonal mean diurnal, seasonal and annual variation
features of the radiation budget components were analyzed in this paper. The results indicated that in spring
diurnal variations of both global solar radiation and the reflective radiation were larger than in other seasons, and
their annual variations were double-peak-shaped, but the phases were different. The distributions of both the
diurnal variation and the annual variation of the earth surface long-wave radiation were unsymmetrical. Annual
variation of the earth effective radiation was of bimodal pattern. One peak corresponded to March and April,
when frozen soil melted, while the other to October, when froze soil froze. Net radiation mainly concentrated in
May, June and July, accounting for 40.14% of the total, indicating that in late spring and early summer the
region's surface had obtained the largest net energy, which played a decisive role for the formation of terrestrial
heat and the heating of the atmosphere.

3. Analysis onpotential evapo-transpiration and dry-wet condition in the seasonal frozen soil region of
northern Tibetan Plateau
This study was based on the observed data at Automatic Weather Site(AWS) of MS3478 in the seasonal
frozen soil region of northern Tibetan plateau from March 2007 to February 2008.The variation characteristics of
potential evapotranspiration (PE) was analyzed based on Penman-Monteith method recommended by FAO. The
contributions of dynamic, thermal and water factors to PE were discussed. Meanwhile, the wet-dry condition of
that region was further studied. The results indicated that daily PE was between 0.52mm and 6.46mm in the
whole year. In summer evaporation was the most intensive, and from May to September monthly PE was over
100mm. In November, there was a clear mutant. Annual potential evapotranspiration was 1037.83mm. In
summer, thermal evapotranspiration was much more significantly than dynamic evapotranspiration; in winter it
was on the contrary. In addition, drought and semi-drought climate lasted for a long time while semi-humid
climate short. The effect of water and dynamic factors on PE varied considerably with the season. Soil moisture
was not the main factor affected PE.

4.Up-scaling scheme was developed
The location of the footprint function varies in time due to changing wind direction and atmospheric stability.
Therefore the footprint of atmospheric measurements does not only affect data quality but also

Page 18 of 98


representativeness of the observed data for the grid level. A scheme to overcome this drawback is in
development and will work in principle as shown in figure 3.

Fig.10. Upscaling scheme for turbulent flux data from heterogeneous landscapes

5.Free convection events at Nam Co site of the Tibetan Plateau were found and analyzed
The spatial and temporal structure in the quality of eddy covariance (EC) measurements at Nam Co site is
analyzed, by using the comprehensive software package TK2 together with a footprint model, and the high
quality turbulent flux data have been obtained for the investigation of free convection events (FCEs). The
research of FCEs at Nam Co site indicates that the generation of FCEs not only can be detected in the morning
hours, when the diurnal circulation system changes its previously prevailed wind direction, but also can be
triggered by the quick variation of heating difference between different types of land use during the daytime
when clouds cover the underlying surface or move away. FCEs at Nam Co site are found to occur frequently,
which can lead to the effective convective release of near ground air masses into the atmosphere boundary layer
(ABL) and may strongly influence its local moisture and temperature profiles and its structure.

Page 19 of 98


Fig. 11. The distribution (a) and frequency statistics (b) of free convection events (FCEs) times at Nam Co site.

6. Diurnal variation of sensible heat flux were very clear
Careful data processing and quality control of LAS has been performed in Naqu BJ station. The comparison
of sensible heat flux measurement by LAS and EC are plotted in Fig12, which shows the similar variation
between LAS and EC measurement.

Page 20 of 98


Fig 12 Comparison of sensible heat flux measurement by LAS and EC (2009.08.01-2009.08.28)

Page 21 of 98


Page 22 of 98


3.2 Work progress in WP2 and achievements during the period
WP2 aims to develop algorithms to retrieve surface parameters from a broad family of multi-spectral and/or
multi-angular radiometric data and produce a consistent data set over the region of Tibetan Plateau.

# Instrument and validation

A multispectral canopy imager (MCI) was developed for the field measurements of forestry canopy LAI. It
can capture image pairs in three different wavelength bands at arbitrary zenithal and horizontal directions.
The MCI image pairs can be used to discriminate the sky, leaves, cloud and woody components. As a result,
this instrument is capable of measuring the woody area index which is very important in field LAI
measurements. In the Heihe river field campaign which was taken in June 2008, MCI was used to get the
directional clumping index and woody-to-total area ratio. Finally, the LAI values were obtained in several
locations after consider the correcting of the clumping effects and woody components.

# Model development

A Whole Growth Stages (WGS) model was developed for simulating the directional reflectance of the row
planted canopy across the whole growth stages. Based on a series of simplifications and assumptions, we
gave out an analytical expression to describe the spatial regular fluctuation of LAVD of row planted wheat
canopy. We found that the LAVD of the vegetal row is approximately negative correlation to the distance
from the centre of the row. Then we put forward a suit of calculation scheme to estimate the directional gap
fraction which well considering the spatial regular fluctuation of LAVD within row-planted wheat canopy. In
our new model, only 4 input parameters are needed, including LAI, the ratio of row width to height, the ratio
of row space to height, row direction.

A new angular & spectral kernel model was developed to describe the BRDF characteristics for most of the
land covers. Compared with the semi-empirical kernel-driven model used by AMBRALS (Algorithm for
Model Bidirectional Reflectance Anisotropies of the Land Surface) which was employed in the MODIS
(Moderate Resolution Imaging Spectra Radiometer) albedo/BRDF product, the component spectra were
combined into the kernel functions instead of kernel coefficients. Then the kernels were expressed as function
of both the observed geometry and wavelength. As a result, the kernel coefficients are independent of
wavelength in this new model. That characteristic enables the broad band conversion to be a linear
combination of the new integral kernels which is much simple and efficient.

A model describing thermal directional radiation was established for the rugged terrain. By parameterization
of sky-view factor and terrain configuration factor, the emitted radiance was parameterized as a linear
composition of the contributions of radiance from vegetation and soil, taking into account the coupling
between vegetation-soil, vegetation-vegetation and soil-vegetation interactive processes.

# A generic inversion algorithm

In order to enable the application of the method to several satellite sensors, the observation model SLC (soil-
leaf-canopy) was extended for applications in the thermal domain, and the MODTRAN interrogation
technique was extended to this domain as well. In addition, look-up table (LUT) techniques were optimized
in order to allow for efficient image simulations under various conditions. This means that for angular
interpolations of the sun-target-sensor geometry only a limited size of the LUTs is required. Topographic
effects were included by considering slope and aspect angles to be obtained from a DEM (digital elevation

Page 23 of 98


model) of the area. Slope and aspect are used to estimate the fractions of solar and sky spectral irradiance in
the optical and thermal domains. A unified equation was derived to describe the TOA radiance as a function
of surface and atmospheric parameters in the optical and thermal domains with the incorporation of
topographic effects. The MODTRAN interrogation technique was extended into the thermal domain as well,
and several MODTRAN outputs were identified with physical quantities of four-stream radiative transfer
theory.

# Topography and scale effect correction for albedo products

One coarse scale pixel includes many tilted micro-areas, which have different slopes and aspects. Its
directional reflectance is affected by these micro-areas and their shadows. An equivalent smooth surface
directional reflectance was introduced for a virtual surface of the coarse scale pixel, which was assumed to be
smooth so that there were no micro-area topography effects. A scale effect correction factor was defined to
correct the topography and scale effect. This factor is only dependent on DEM and the geometry of sun and
sensor. The topography and scale effect correction algorithm includes three steps: (1) Setting up a database for
pixel-average slope and aspect angle for each pixel of 500m grid and 5km grid, and scale effect correction
factor for each 5km pixel; (2) Correcting the pixel level topography effect for 500m directional reflectance,
using slope and aspect angles; (3) Correcting the pixel level, as well as subpixel level, topography effect for
5km directional reflectance, using slope, aspect angles and the scale effect correction factor.

# A priori knowledge based LAI inversion

The a priori knowledge of LAI was obtained by three ways: (1) Getting the relationship between a
multidirectional averaged NDVI and LAI by simulation using a BRDF model (eg. SAILH model); (2)
Developing the empirical crop growth model by the regression of a LOGISTIC equation and the field
measured LAI data sets; (3) Developing a priori LAI trend from several years’ MODIS LAI product. All of
this a priori information was used in the inversion of radiative transfer models to get the temporal continuous
and robust LAI. Both of the MODIS and MISR data were used in the inversion to improve LAI product.

# Angular effect correction of fractional vegetation cover

Under the assumption of that a remote sensing pixel is mixed by vegetation and background, a simple
directional fractional vegetation cover (FVC) model was developed based on Beer-Lambert law. The
variables in this model can be got by using the MODIS images in 16 days and high resolution HJ-1 images
The Scaled Trust-Region Solver for Constrained Nonlinear Equations (STRSCNE) algorithm was used to
retrieve the variables. A vegetation growth model was introduced to constrain the relative worse quality of HJ
data in a temporal scale. The different spectral responds of MODIS and HJ were also compared with
spectrums of typical surface class. Uncertainty was assessed by error propagation theory and field
experiments.

# LST inversion using polar satellite data

A review of existing algorithms to retrieve land surface emissivities (LSE) and land surface temperatures
(LST) has been carried out. This review has allowed the selection of the needed algorithms to retrieve LSE
and LST, which includes the preliminary determination of several parameters such as NDVI (Normalized
Difference Vegetation Index), FVC (Fraction of Vegetation Cover), total atmospheric water vapour content,
as well as carrying out cloud tests, image atmospheric and geometric correction. In the absence of the MODIS
– CEOP-AEGIS dataset, these algorithms are being implemented on the data acquired by the Global Change
Unit at the University of Valencia (Spain), in order to obtain a near-real estimation of LSE and LST. The
completion of this process is expected during the next reporting period. In a second step, this processing chain
will be adapted to the Tibet area in order to process the MODIS – CEOP-AEGIS dataset.

Page 24 of 98


The algorithm of daytime 150m LST product was proposed by using the HJ-1 dataset over the Tibet Plateau.
A view angle dependent single channel LST algorithm has been developed for correcting atmospheric and
emissivity effects for all land cover types.

• Highlight clearly significant results (3 pages)

# Multispectral canopy imager (MCI) and its use in woody-to-total area ratio determination

The MCI, which mainly comprises a near-infrared band camera, two visible band cameras, filters and a pan
tilt, was developed to measure clumping index, woody-to-total area ratio and geometrical parameters of
isolated trees (figure 1). Two typical sampling plots (Plots 1 and 5) which were covered by Picea crassifolia
were selected for the estimation of woody-to-total area ratio and its directional change in Heihe river basin,
China. The clumping index and woody-to-total area ratio values of the forest canopy were got at eight zenith
angles (from 0 to 70° in increments of 10°) using MCI images based on gap size distribution theory (figure
2,3).

Figure 1. Illustration of the multispectral canopy imager (MCI).

Erreur ! Des objets ne peuvent pas être créés à partir des codes de champs de mise en forme.Erreur ! Des objets ne
peuvent pas être créés à partir des codes de champs de mise en forme.
Figure2. Clumping indices at Plot 1 (a) and Plot 5 (b).

Erreur ! Des objets ne peuvent pas être créés à partir des codes de champs de mise en forme.Erreur ! Des objets ne
peuvent pas être créés à partir des codes de champs de mise en forme.
Figure3. The woody-to-total area ratio of Plot 1 (a) and Plot 5 (b).

The detailed description of the equipment and the method can be found in the following paper:

Page 25 of 98


Jie Zou, Guangjian Yan, Lin Zhu and Wuming Zhang, Woody-to-total area ratio determination with a
multispectral canopy imager (MCI), Tree Physiology, 2009; doi: 10.1093/treephys/tpp042.

# Unified modelling of TOA radiance for the generic inversion algorithm

A unified equation was derived to describe the TOA radiance as a function of surface and atmospheric
parameters in the optical and thermal domains with the incorporation of topographic effects. This equation
reads:

where and are the viewing factors associated with illumination from the sun and the sky, respectively.
They are given by

,

where and are terrain slope and aspect, respectively.

The four terms in square brackets are the ones associated with:

• Atmospheric path radiance in both domains
• Adjacency effects in both domains
• Sky irradiance contributions in both domains for the target
• Direct solar bi-directional and thermal direct target contributions

Note, that emissivities are represented here by their associate reflectance equivalents and
(hemispherical and directional emissivity).

# Time series LAI mapping over Heihe river basin

Page 26 of 98


The developed variational assimilation method was implemented and some results on LAI values for the
whole year of 2008 over Heihe River Basin are presented in Figure 4. It shows the regional LAI mapping
results from the time series MODIS reflectance data acquired over this area in 2008 with the spatial
resolution of 500m. As seen, temporal variation of the LAI values in this region is reasonable. And the spatial
variability is consistent with the vegetation cover map in this area.

Figure 4 LAI inversion results in the middle of Heihe River area.

# Emissivity measurements and data preparation

Several papers have been published regarding different topics of LST from polar satellites such as:

(1) José A. Sobrino, Cristian Mattar, Pablo Pardo, Juan C. Jiménez-Muñoz, Simon J. Hook, Alice Baldridge,
and Rafael Ibañez. 2009. Soil emissivity and reflectance spectra measurements. Applied Optics, Vol. 48, Issue
19, pp. 3664-3670.

This work present a laboratory procedure to characterize the emissivity spectra about several soil samples
collected in diverse suite of test sites in Europe, North Africa, and South America from 2002 to 2008. Here,
we presented a cross calibration with in-situ measurements and further application to thermal remote sensing.
This work presents a methodology to characterise the emissivity values of a given soil sample, additionally,
the soil emissivity values analyzed here were presented for all polar satellites which have thermal sensors.

(2) C. Mattar, J.A. Sobrino, Y.Julien, J.C. Jiménez-Muñoz, G. Soriá, J. Cuenca, M. Romaguera, V. Hidalgo,
B. Franch, R. Oltra. 2009. Database of atmospheric profiles over Europe for correction of Landsat thermal
data. Proceedings of the 33rd International Symposium on Remote Sensing of Environment. (in press)

This work presents a new vertical profile data base for correct thermal remote sensing images. In this case we
focused our work to provide useful information to correct Landsat thermal images. However, the data base
could be used for other remote sensing sensors.

# Spectra normalization of HJ and MODIS data
Difference of spectral responds of HJ and MODIS sensors should be considered in FVC retrieval, though
MODIS and HJ sensors have overlapped region in spectral respond functions (figure 5). Many reflectance
spectrums of leaves and soils were selected from spectrum library of ENVI software. The mean values were

Page 27 of 98


computed for the two sensors (table 1). Scattering plot of 4 bands in figure 6 didn’t exhibit much difference
for HJ and MODIS.

Figure 5. Relative spectral respond function of MODIS and HJ-1 bands used in FVC retrieval

Table 1. Mean reflectance of typical land covers with HJ and MODIS relative spectral response
Reflectance of typical leaves and soils
conifer deciduous Grass and soil
arbre
Blue HJ-1 0.0704562 0.07849 0.08478 0.077605
MODIS 0.0621984 0.065187 0.071822 0.064877
Green HJ-1 0.100901 0.132595 0.135229 0.139566
MODIS 0.114949 0.149223 0.14475 0.12815
Red HJ-1 0.075 0.119595 0.129705 0.204328
MODIS 0.071389 0.110964 0.12425 0.195855
Near- HJ-1 0.51273 0.683053 0.517343 0.281649
infrared MODIS 0.525689 0.692068 0.534383 0.300353

Scattering plot of reflectances

Blue
Green
Red
NIR

Figure 6. Reflectances of HJ-1 and MODIS signal corresponding to typical land cover types

# Development of a quantitative remote sensing products inversion system

Page 28 of 98


A quantitative remote sensing products inversion system is being developed for the parameters products
generation. It is composed of 5 sub-systems, including database, data pre-processing, products inversion,
validation, and visualization.
(1) Database subsystem takes charge of the data management and data flow of the whole system. All the
other sub-systems will be connected together by database without physical connection between the 4
sub-systems;
(2) Data pre-processing subsystem will process all the incoming remotely sensed data into standard data
products. The pre-processing procedures include cross radiometric calibration, geometric correction,
projection transferring, gridding, and cloud screening;
(3) Products inversion subsystem is a products “pool” which is composed of 22 geo and bio parameters and
system users will make their own product producing workflow. The subsystem will be producing
products through the workflow instantaneously or routinely;
(4) Validation subsystem will validate the inversion products based on the predefined methods routinely or
by users’ convenience;
(5) Visualization subsystem is a visual interface which provides users with data management, image display
environment, image and graphic processing, terrain analysis, statistics analysis, and annotating.

Page 29 of 98


Page 30 of 98


Summary of progress towards objectives, per task:

Task 3.1 (ALTERRA, ITC, BNU, CAREERI, TUD): local validation of algorithms with ground eddy
covariance measurements at footprint scale and cross-comparison of approaches to turbulent flux partitioning.

The remote sensing based algorithm for flux calculation to be evaluated in this task can be applied at local
scale (S-SEBI, SEBS) or at a larger (meso) scale (SEBS, MSSEBS). They all follow the approach proposed
by Menenti and Choudhury (1993) stating that for a given net radiation value, and for homogeneous
atmospheric conditions, the surface temperature is related to the ratio between actual and potential
evaporation. Both methods require physical properties of the surface extracted from remote sensing to
characterize the surface radiative balance (albedo, surface temperature, emissivity) and vegetation structure
(fractional cover, Leaf Area Index). Also they differ in the way to define wet and dry boundaries in terms of
normalized surface to air temperature gradient, they all require some basic meteorological information.
Therefore the contribution of UDS in this task consisted in:
i. identify remote sensing products available to conduct SEB calculation for areas and periods of time
where reliable ground measurement data were available;
ii. gather and post-process meteorological data to be used as forcing conditions in the SEB schemes

The remote sensing products used to conduct the algorithm comparisons are Modis images acquired by Terra.
The reasons are: i. the adequate spatial and temporal resolution of the sensor; ii. the panel of adequate
products; iii. ad hoc products from WP2 are not available at this stage of the project. The products and dates
are summarized in the tables bellow. The candidate dates were selected on the basis of global cloudiness on
the Plateau.

April 2003 15th and 25th
May 2003 28th
October 2003 17th and 23rd
November 2003 8th and 11th

Product Variable Spatial resolution Temporal resolution
MOD11A1 LST/Emissivity 1km Daily
MCD43B3 Albedo 1km 16 days
MOD13A2 Vegetation index NDVI 1km 16 days
MOD15A2 LAI 1km 8 days

The characterization of the state of the Planetary Boundary Layer is based on the output from the Meso-scale
Numerical Weather Prediction Model GRAPES developed by the Chinese Academy of Meteorological
Sciences, partner in this project. The following variables were extracted from GRAPES simulations covering
the entire Plateau at a resolution of 30 km and 30’ time step:
Variables extracted at the height of the Atmospheric Boundary Layer:
• ABL height
• Air temperature
• Specific humidity
• Wind speed
• Air pressure
Variables needed at 2 meters:
• Specific humidity

Page 31 of 98


• Air pressure

UDS prepared two set of inputs centered on the validation site called BJ, with either 400x400 km or 100x100
km extent. The processing consisted in:
• extraction of MODIS products, re-projection and spatial re-sampling of albedo, LST with
corresponding acquisition time layer, NDVI, LAI
• extraction of GRAPES outputs from GrADS raw files, geo-processing of layer variables to the same
resolution and coverage as MODIS products
• creation of time-composite PBL layers to associate adequate GRAPES field to MODIS LST
following MODIS LST acquisition time
• extraction of SRTM Digital Elevation Data for the selected scene to calculate PBL elevation

This dataset was used to perform S-SEBI, SEBS and MSSEBS calculations, tests and comparisons (see next
section).

Task 3.2 (UDS, ALTERRA, ITC, ITP, BNU): generalize SEB calculation at a high spatial resolution and on a
regional extent. On such an extent, local towers cannot be used to define boundary conditions. The MSSEBS
(Colin, 2006) approach enables to link ground variables at a high spatial resolution (typically 30 meters) with
Atmospheric Boundary Layer (ABL) state at a proper resolution related to the typical ABL length scale.

Generalize SEB calculation on the entire Plateau lead to several conceptual and technical challenges:
• the combination of high resolution remote sensing products with medium (meso) resolution NWPM
outputs in a single calculation scheme, combining physical variables whose meaning is closely
related to their inherent scale, as to be taken into account in the algorithm implementation
• the use of high (1km) resolution remote sensing products over the Plateau lead to significant amount
of data (e.g. 1,400 x 1,700 km grid means 2.4E6 calculation nodes, for n variables and j time steps
with n > 25 and j >> 100).
• the use of NWPM with different spatial and temporal resolution, geo projection, supposes to have a
powerful pre-processing procedure to mix various data sources in a single model input set of layers
• the probable occurrence of data unavailability (clouds…), data inconsistency (NaN, error code)
supposes to have a flexible enough implementation to manage with various situations with a
minimum of manual work
These considerations lead to the prototyping and current development of a new SEB framework, with the
following characteristics:
• core algorithms are separated from I/O procedures; external I/O procedures can be extended without
any modification of the algorithms to allow the use of new data sources
• efficient object oriented python coding based on Numpy and SciPy math libraries for fast processing
of numerical arrays; multi-core computation capability; fully open source based and cross-plateform
• XML based configuration, with HTML/PhP user interface (under development)
• powerful geo-processing library GDAL embedded
• self-diagnosis capability for fast analysis of mass of log files
At this stage of the project, this code is under development, with evaluation of a beta version. The first stable
version will be described in details in the Algorithm Theoretical Basis Document to be delivered on
milestone M2. The resulting products will be made available to WP8 partners, and as a new product in the
database of the project to be registered to GEOSS.

Page 32 of 98


Figure 1: SEB framework chart

Task 3.3 (UDS, ALTERRA, ITC, ITP) : The same MSSEBS approach is used with low resolution satellite
images (Feng Yun-2) and NWP model outputs over the entire plateau. These low resolution fluxes maps can
be validated from spatially integrated maps obtained in Task 3.2.
(nothing at this stage of the project)

Significant results

The aim of the calculations performed with the 2003 dataset is to perform a cross-comparison of algorithms
and a validation with ground measurements.
The candidate algorithm of UDS is the Multi-Scale Surface Energy Balance System (Colin, 2006). This is a
single source SEBI based scheme designed to process radiative balance, PBL stability and external
resistances at appropriate scales as regards the physical meaning of key variables (e.g. roughness length for
momentum and heat, stability functions in the atmospheric boundary layer…), to produce evaporative
fraction maps. The soil heat flux is computed following vegetation fraction data, and the total diurnal
evaporation is computed with a locally fitted model of net available energy for turbulent flux. The sensible
heat flux is calculated as the residual of the energy balance.

Page 33 of 98


Figure 2: example of results for Nov 11th 2003: (top left) PBL forcing from GRAPES, values are allocated following the
acquisition time of the LST; (top right) MODIS products; (bottom left) Sensible heat flux map from MSSEBS; (top
right) Latent heat flux map from MSSEBS.

For the 1x1 km pixel where the Bijie site is located is, for Nov. 11th 2003 at 11:06, the latent heat flux calculated
with MSSEBS is 7.6 W.m-2 , and the sensible heat flux is 143.1 W.m-2, while ground values of latent heat flux
measured at respectively 10:30 and 11:30 range from -14.3 W.m-2 to -55.5 W.m-2, and the corresponding sensible
heat flux ranges from 91.4 W.m-2 to 200.0 W.m-2.
Since the latent heat flux from MSSEBS is of the order of magnitude of the model uncertainty (Colin 2006), the
evaporation can be considered as almost negligible. Moreover, as the ground measurement values used here are
sensor values, a comparison with a 1 km resolution pixel would require further analysis of the spatial meaning of
the measures.

This first experiment gives important information for the preparation of the next phase of the project:
• whatever the date of the year, even a limited scene is affected by clouds. The SEB framework has to be
able to deal with missing values in mathematical processing, and gap filling technics to be implemented
in WP2 will probably be critical to provide a continuous flow of inputs for the time series processing
phase to come

Page 34 of 98


• also these experiments are based of GRAPES simulations, GRAPES usually provide analysis data, ie. at
a fixed 6 hour time step. This is of consequence as regard the acquisition time of LST products. An
additional step may be required to derive LST at a GRAPES time step from the remote sensing products.

This first experiment has several significant limitations:
• no data were available to conduct a dual-source calculation
• validation data were only available for one point, and local meteorological conditions only allowed to
use one of the selected dates
• ground measurement data used for validation didn’t passed through detailed quality and footprint
analysis

Therefore a new validation experiment was initiated with a selection of 3 different sites located in very different
parts of the Plateau, using 4 sets of 10 days of data in January, April, July and October 2008. This set of
validation data was made available late September 2009 by WP1 partners. MODIS products were collected, and
GRAPES simulations still have to be performed at the time of writing this report. Therefore it is asked that the
target delivery time of deliverable de 3.1 “Review of selected existing algorithms and models on local, regional
and Plateau scales data sets” is set to December 20th to allow for the completition of this analysis.

Page 35 of 98


Summary of progress

Task 4.1: Review and inter-comparison of available algorithms and products (microwave backscattering
coefficient, microwave emissivity and land surface temperature diurnal cycle) (ITC, CAREERI, BNU,
IGSNRR)
This task is completed and report is written

Task 4.2: Collection of consistent continuous in-situ soil moisture measurements at regional scale of
selected sites on the Tibetan plateau measurements which will include soil moisture (including soil
temperature, vegetation parameter, soil texture and land surface roughness) at two sites (Maqu-grassland,
and Naqu tentatively) (CAREERI, ITC)
Task 4.2 has been completed and Deliverable 4.1 has been distributed.
CARRERI and ITC have installed in May-July 2008 an extensive soil moisture and soil temperature
monitoring network in the water source region of the Yellow River to the South of Maqu city, on the border
between Gansu and Sichuan province, in China (33°30’-34°15’N, 101°38’-102°45’E). The network consists
of 20 stations monitoring the soil moisture and temperature at different depths (from 5 to 80 cm deep) every
15 minutes. The network covers an area of approximately 40 km*80 km, where the elevation ranges between
3430 m and 3750 m a.s.l (north-eastern edge of the Tibetan Plateau). To ensure complete data continuity, the
data are downloaded twice per year by CAREERI: at the beginning of the monsoon season (in May) and at
the end (in October).
A specific calibration of the probes has been carried out for the soil type of Maqu area, increasing the
accuracy of the soil moisture measurements from 6% to 2%.
The quality of the data downloaded from Maqu monitoring network has been checked by evaluating their
consistency in time and space and by comparing their trends with meteorological data and with soil moisture
satellite products. A clear consistency and a good agreement have been found.
The calibrated data collected at all the stations and at all available depths are reported in an Excel file and a
detailed technical report has been attached to the data. Both of them have been delivered to the project teams.

Task 4.3: Development of a satellite sensor independent system for the soil moisture combined retrieval
algorithms (ITC, CAREERI, BNU)
This task is in progress. A retrieval model is developed for ASCAT data which will be combined with passive
microwave data in the course of the project.

Task 4.4: Estimation of soil moisture from Geostationary Satellite (GS) data (optical remotely sensed
data) (IGSNRR)
In order to develop method of estimate soil moisture based on geostationary satellite data using the diurnal
variation of LST derived and global radiation (shortwave). Following investigations were conducted during this
time:
1. Construction of land surface diurnal temperature cycle model and the ellipse relationship between
LST and solar shortwave radiance.
In geostationary satellite observation system, there are adequate images to describe land surface temperature
variation under clear sky condition. In generally, land surface temperature diurnal variation can be expressed as a
harmonic term in daytime and an exponential term during the nighttime. This two-part semi-empirical diurnal
temperature cycle (DTC) model has used by Göttsche and Olesen (2001), Schädlich et al. (2001) and Jiang et al.
(2006). In our work, we chose the model applied in Jiang (2006).

2. Land surface temperature simulation with land surface model (i.e. Common Land Model )

Page 36 of 98


In order to validate some assumption and analyze the method mentioned above, simulation data is an easy and
fast way. In our simulation, Common Land Model was selected to simulate land surface temperature under
different environment conditions in clear air condition. During the simulation, soil type and land cover type were
usually set to be constant. Then we modeled the land surface temperature variation under different percent
vegetation cover varying from 0%- 100% with a step of 10% and soil volumetric water content varying from
0%-50% with a step of 5%.
Several parameters were extracted from the land surface temperature daily cycle like maximum temperature,
minimum temperature, daily temperature amplitude, temperature morning raising rate and so on. Correlation
analysis was conducted here to analyze the relationship of there parameters with soil water content and percent
vegetation cover. The results showed that land surface temperature is a complex variable. It is influenced not
only by soil water content, but also is greatly influenced by surface land cover type and percent vegetation cover.
As an interface between land and air, Land surface has strong energy and material exchange processes. In order
to understand the degree of soil water content’s influence on land surface temperature, the other factors should
be eliminated firstly.

3. Organization and implement field experiment in Lang fang experimental base.
Beside land surface model simulation, we also organized a field experiment in Lang fang experimental base in
He bei province, China. In order to measure the atmosphere and soil data, such as air temperature, wind velocity,
soil volumetric water content, we purchased an Automatic Weather Station and Time Domain Relectometers
(TDR). Meanwhile, land surface temperature was measured by infrared thermometer. Down-welling globe
radiation and net radiation were also recorded using Solar Radiometer.
The experiment was implemented from 17th Oct. to 5th Nov. 2008 for 20 days. Three sites were executed
simultaneously with three soil types (sand, watered local soil and non-watered local soil).

4. In-situ measurement data analysis
From the experiment, many data was collected. Fig. 3.4.1 shows the observed records of soil surface
temperature, wind speed and air temperature at 2 Meter height of 5 days.

Fig.3.4.1 Sample of observed data during the experiment

Page 37 of 98


From the observed data, we analyzed the soil temperature raising rate related to the Net Surface Shortwave
Radiation (NSSR) during the morning time, and the temperature falling rate related to NSSR or Net Surface
Radiation (NSR) during the afternoon time.

5. Abnormal surface nocturnal cooling effect analysis
From the in-situ measurements and satellite data of MSG SEVIRI, we found that there exists an abnormal rising
of the change of the soil temperature in the nocturnal cooling process. Nocturnal surface intense cooling may
result in the inversion of the atmospheric temperature and water vapor. In order to study the abnormal
phenomenon, we analyze and simulate the changes of surface temperature under different atmospheric
conditions

Task 4.5: A data product of the plateau using different sensors simultaneously (AMSR-E, ASCAT,
SMOS) (BNU, ITC)
Up to October 2009, we had collected all of the satellite observation data and ancillary data used for retrieval,
including AMSR-E Level 2A, Level 3 brightness temperature data, SRTM 90m DEM data, MODIS IGBP
land cover map, and surface freeze/thaw status data, etc.
Available ground surface emission models were evaluated and compared in detail, on this basis, a forward
simulation system was established. It uses Qp model to calculate the emission of rough soil surface, and !-"
model to consider the vegetation effects. Through simulation analysis, the crucial inversion methods were
determined. A multi-channel temperature estimation algorithm using AMSR-E was selected to obtain the
surface temperature. The new developed microwave vegetation Indices (MVIs) was used to eliminate the
vegetation effects. And a soil moisture index developed from Qp model was put forward to minimize the
effects of surface roughness. When the above methods were used in the soil moisture retrieval, some good
results were achieved, and further results are still in progress.

Task 4.6: Validation results and documentation of uncertainties (CAREERI, BNU)
There is no progress made so far and is in accordance with project plan.

Significant results

Collection of consistent continuous in-situ soil moisture measurements at regional scale
One of the objectives of the CEOP-AEGIS project is to develop a soil moisture retrieval algorithm based on the
simultaneous use of active and passive microwave satellite data. The developed algorithm is sensor
configuration independent and is able to incorporate data of present and future satellite data, such as AMSR-E,
ASCAT and SMOS. The long term and large scale products obtained applying the developed algorithm over the
Tibetan Plateau, will be extremely important to understand the links between Monsoon system, precipitation
patterns and soil moisture.

For this reason, extensive soil moisture monitoring networks are required to obtain ground information which
can be compared to the retrieved soil moisture products, in order to evaluate their consistency.
To tackle this validation problem, CARRERI and ITC have installed in July 2008 an extensive soil moisture and
soil temperature monitoring network in the water source region of the Yellow River to the South of Maqu city
(Gansu province, China). The network consists of 20 stations monitoring the soil moisture and temperature at
different depths (from 5 to 80 cm deep) every 15 minutes. The network covers an area of approximately 40
km*80 km.
The area selected for the installation of an extensive soil moisture monitoring network is located to the South of
Maqu city, on the border between Gansu and Sichuan province, in China. The network is at the north-eastern
edge of the Tibetan Plateau (33°30 -34°15’N, 101°38’-102°45’E) and at the first major meander of the Yellow
River, where it meets the Black river. It covers the large valley of the river and the surrounding hills (Figure

Page 38 of 98


3.4.2), characterised by a uniform land cover of short grassland used for grazing by sheep and yaks. In this area
the elevation ranges between 3430 m and 3750 m a.s.l.

The installation of the soil moisture and soil temperature monitoring stations started in May 2008 with the
stations CST_01-05 and was concluded at the end of June 2008 with all the other stations. Therefore since July
2008 the complete network is operative.
The network covers an area of approximately 80 km*40 km and the locations have been selected in order to
monitor the area extensively at different altitudes and for different soil characteristics.

During the installation, soil samples were collected in order to analyse bulk density, particle size distribution and
organic matter content. The samples for particle size and organic matter were collected at a depth between 5 and
15 cm. A laser scanner (Mastersizer S Ver. 2.18 by Malvern Instruments Ltd.) was employed to estimate the soil
particle size distribution and the standard method for the organic matter content. Soil sample rings (aluminium
cylinders of known volume) were collected at 5 cm depth and oven dried at 105°C to estimate the bulk density
(i.e. dry soil mass in a known volume). When the soil profile showed a variation at deeper layers, the sample
collection and the analyses were repeated for the second horizon as well.

Figure 3.4.2 Maqu area, Yellow River valley and location of the 20 soil moisture and soil temperature stations of the
network.

Each network station consists of one Em50 ECH2O datalogger (by Decagon), which is recording the data
collected by two to five EC-TM ECH2O probes (by Decagon) able to measure both soil moisture and soil
temperature.
EC-TM ECH2O probe consists of 3 flat pins 5.2 cm long. It is a capacitance sensor measuring the dielectric
permittivity of the soil surrounding the pins. The dielectric permittivity is then converted in volumetric soil
moisture according to a standard calibration equation. The soil temperature is measured using a thermistor
located on the same probe.

Page 39 of 98


Figure 3.4.3 Installation procedure

A specific calibration of the probes was needed for the soil type of Maqu area. Therefore soil samples were
collected and laboratory calibrations were carried out (see following paragraph).
For the installation, a deep hole in the soil was dug and the probes were installed on one of the hole walls, at
different depths and with the pins in horizontal direction. Then probes and datalogger (closed in a box) were
completely buried (see Figure 3.4.3).

EC-TM ECH2O probes estimate the volumetric water content of the soil by measuring the dielectric constant of
the soil. However the dielectric properties of the soils depend on soil texture and salinity. Decagon has
determined a generic calibration equation (applied by default by the datalogger), which is valid for all fine
textured mineral soils with an accuracy of approximately ± 3%. This accuracy can be increased to 1-2%,
performing a soil-specific calibration. For this reason about 5-6 kg of soil were collected in each location at a
depth of about 5-15 cm (as well as at deeper layers, in case the soil profile was different) in order to carry out a
laboratory specific calibration, following the instruction guide provided by Decagon.

Figure 3.4.4 Results of the soil specific calibration of ECH2O probes

Page 40 of 98


In conclusion, the calibration (Fig.3.4.4) has led to a decrease of the rmse between the volumetric soil moisture
measured by the rings and that measured by the probes from 0.06 to 0.02 m3/m3.

Page 41 of 98



Estimation of precipitation over the Plateau and surrounding zones with optical and microwave
observations

The objective of this WP was twofold: to provide multisensor and multiplatform observation of precipitation
over the Plateau, and to get a deeper understanding of cloud and precipitation processes ongoing over this area.
The temporal development of the activities identified as the first step the set-up of a reliable strategy to provide
quantitative precipitation measurement. This was achieved during the first reporting period of the project: the
weather radar data have been pre-processed to provide the project with a quality controlled 3D precipitation
dataset over the project target area.
On the other side, two studies were completed indicating that prevailing synoptic scale trough is one of a key
indicator to establish unique precipitation system over the Tibetan Plateau. Other activities are in their first
developing phase, and did not yet achieved significant results, as planned in the DoW document.
In the next pages a more detailed description of the activities is presented task by task.

Summary of progress.

Task 5.1: To observe the cloud and precipitation microphysics processes in Tibetan Plateau and
southwestern China by cloud Doppler radar, movable X and C band dual linear polarization radar. A
hydrometeors classification algorithm will be applied to retrieve the 3D microphysical cloud structure.
The radar observation has started in the sites operated by CAMS: the radar network and rain gauge information,
analyze the ground blockage for radar in Tibetan and Qinghai Province. The results show that the radars in
Tibetan are blockage by around mountain severely, the radar coverage is limited. The radar in Qinghai province
can be used to precipitation estimation with rain gauges. A fuzzy-logic based algorithm for hydrometeor phase
classification with polarimetric radar has been developed by CAMS. A small network of three X-band
disdrometers (PLUDIX) is planned by UNIFE (with the assistance of ITP-CAS) and the installation will be
completed in November 2009.

Task 5.2: To develop the QC and mosaic algorithms for operational Doppler radar network. The
disdrometric data will be used in radar QC and for radar calibration if disdrometers instruments are
available.
Research work on radar data quality and reflectivity remap and mosaic has been carried on by CAMS, and the
algorithm for 3 D mosaic. A the fuzzy logic based algorithm is used to detect the anomalous propagation and
ground clutter; four interpolation approaches are used to remap raw radar reflectivity fields onto a 3D Cartesian
grid with high resolution, and three approaches of combining multiple-radar reflectivity fields are used. The
algorithm has been used to process the radar data and provide 3D data to the other partners of WP5.
In particular, the raw precipitation data in Tibetan and the gridded precipitation data were provided by CAMS to
UNIFE for two case studies. for period of 18 June 2008-19 June 2008 and 18 July 2008-20 July 2008, with
spatial and temporal resolution (0.01°#0.01°#0.5km#5min)
Finally, CAMS processed radar data and provided 3D reflectivity data to WP5. Grid Reflectivity in Qinghai
from 18 July 2008 -21 July 2008 were product, the radar data in Tibetan from 18 June 2008 to19 June 2008, 18
July 2008 to 20 July 2008 were provide.
The data of three X-band disdrometers will made available by UNIFE for the period 1 November 2009 – 30
October 2010, to improve the quantitative radar rainfall products.

Page 42 of 98


Task 5.3: To analyze the meso-scale structures and processes of precipitation systems in Tibetan Plateau and
southwestern China by operational Doppler radar network in China and satellite (e.g. cloud products of MODIS).
The precipitation distributions with different algorithms will be compared in case studies.
UNIFE carried out an inventory of satellite precipitation estimation techniques, including both physical and
statistical approach and considering microwave (AMSR-E, SSM/I-SSMIS and AMSU), visible-infrared (MODIS,
AVHRR, Meteosat, FY-2C), and blended techniques. The characteristics of different techniques were analyzed
to select the more suitable ones for application over the Tibetan Plateau. The events proposed by CAMS were
selected as case study for the early application of selected techniques.
UNITSUK completed an analysis of the meso-scale structures and processes of precipitation systems and
identification of the indicators for the rainfall processes in Tibetan Plateau (TP) and southwestern China, and the
results will be summarized in the next section.

Task 5.4: To use the rain maps obtained by the ANN technique along two main lines: improve the performance
of floods and drought warning systems, and analyze long term (seasonal) rainfall pattern.
IGSNRR performed an inventory of Satellite Rainfall Estimation approaches and studied the theory of Artificial
Neural Network (ANN) and application in satellite rainfall estimation. The MATLAB software is considered for
ANN implementation. A first satellite dataset (June 2007 to September 2007) has collected and processed: FY-
2C satellite images (provided by the National Satellite Meteorological Center of China at 5Km spatial resolution
and hourly temporal resolution) and Gauge data (purchasing from the National Satellite Meteorological Center
of China) at hourly temporal resolution as well.
An ANN technique is implemented and tested with gauges data by IGSNRR, and the preliminary results will be
summarized in the next section.
UNIFE started to apply an ANN technique developed for MODIS data and focused on mid-latitude, to the case
studies over the Tibetan Plateau.

Task 5.5: To retrieve the precipitation with Doppler radar, satellite data and rain gauges in mountain region.
The retrieval of precipitation fields from radar and rain gauges has started (see task 5.2), while the satellite
approach is still in its preliminary phase (see also Tasks 5.3, 5.4 and 5.8).

Task 5.6: To obtain the distribution of Precipitable Water Vapor (PWV) in Tibetan Plateau and its
adjoining area by GPS receiver.
This task is not yet started by CAMS.

Task 5.7: To obtain the indicators of the rainfall process in Tibetan Plateau and southwestern China by analyzing
the change of PWV.
UNITSUK carried on a study on the relevance of water vapor transportation processes, using reanalysis data and
numerical weather prediction output. Results of this study will be summarized in the next section.

Task 5.8: To improve the current combined precipitation estimation technique with the radiometer
(TMI) and PR with the simulation database developed above and inclusion of the effects of topography over the
Plateau; Also here we will correct the satellite estimation of precipitation using the ground rain gauge data in the
algorithm, and validate the inversion scheme with ground observation.
For this task UNIFE planned to apply a rainfall retrieval scheme that works on conical scanner data (SSM/I-
SSMIS, AMSR-E, TMI). The algorithm is based on a cloud radiation database constructed as follows. A cloud
profile data set is assembled by means of cloud resolving model outputs (the Non-hydrostatic Modeling System
of the University of Wisconsin is used to this end), then a radiative transfer algorithm is applied to simulate the
radiances upwelling from the modeled cloud profiles. When a set of satellite radiances is measured from a given
sensor, the database is searched for the cloud profile whose simulated radiance better match the observed ones.
This algorithm is currently applied in different regions with encouraging results.

Significant results

Page 43 of 98


2.1 Delivery of radar 3D precipitation gridded data (CAMS)
A remarkable result of the first 18 months of WP5 activity is the production of a radar derived precipitation
product gridded on a 3D grid (mosaic). The used techniques and the developed algorithms are described in the
deliverable (D5.1) about radar data pre-processing issued by CAMS, responsible for the radar data.

2.2 Studies on moist processes over Tibetan Plateau (UNITSUK)
To reach the general WP5 objective of improving the understanding of cloud, water vapor, and precipitation
processes of mountain area in Tibetan Plateau (TP) and Southwestern China, two relevant results were achieved
by the analysis of the meso-scale structures and processes of precipitation systems and identification of the
indicators for the rainfall processes in TP.

At first, water vapor (WV) transportation processes into the TP during monsoon season was examined by
reanalysis data and numerical simulation, and found that synoptic scale troughs separated by the TP played
primarily rules to intrude the mid-troposphere WV and converge over the southeast on the TP. The systematic
intrusion occurred at the same time with Indian monsoon breaks. Numerical simulation also indicated that
daytime valley winds play secondary function to bring WV into higher elevations over the southern slopes of the
TP even prevailing with passing of troughs. Those evidences need to be verified by in-situ observation data that
are planned to be archived by the WP5 partners in the later stages. The results are published in the Journal of
Meteorological society of Japan by Sugimoto et al. (2008).

On the TP, there are two types of convection processes, one is the thermal convective activities during daytime
and the other is the stratified nocturnal system. During the first stage, we are more focused on the analysis of
nighttime precipitation system. Frequent occurrence of the nocturnal precipitation was often reported and
identified as a key parameter to control morning soil moisture amount which could feedback the next day’s
starting of convections. However, there were few studies treated the mechanism of nighttime precipitation on
the TP. Composite analysis of the in-situ observation data and re-analysis data showed that precipitation after
the mi-night frequently occurred with easterly winds provided by an anti-cyclone located in the north-west of the
TP through successive days. Numerical simulation revealed that the anti-cyclone was formed by the plateau
scale topographic effect with a traveling of mid-latitudes baroclinic wave, which caused synoptic scale
convergence zone over the central PT and activate convections by dissolving near-surface convective instability.
It can be explained that such convergence would be dissipated during daytime due to prevailing of strong
thermal convections and associated sub-grid scale local circulations. The results are published in the Journal of
Meteorological society of Japan by Ueno et al. (2009).

Those two studies indicated that prevailing synoptic scale trough is one of a key indicator to establish unique
precipitation system over the TP. Accurate prediction of the location and development of troughs with its year-
to-year variability would be required to assess long-term water cycle trends. Research activities regarding to the
development of meso-scale convective systems, that links to severe weather prediction, have been conducted as
a part of WP7 and explained there.

2.3 ANN implementation to retrieve precipitation from FY-2C data (IGSNRR)
Preliminary results were also reached with the use of an Artificial Neural Network (ANN) rainfall estimation
technique. Several ANN types have been considered for application: in this work a three-layer feedforward
neural network (TLFNN) was implemented. Output layer has only one neuron since the objective is to estimate
the precipitation value associated to a certain pixel(i,j), otherwise, a log-sigmoid and “purelin” activation
function is used for the first hidden layer and the second hidden layer respectively. In table 1 the input of the
TLFNN are listed.

Lots of researches show that Tibetan Plateau was mainly rainy in summer, so in this work we focused on
summer season. The performance of a NN depend on the choice of network model, architecture and parameters

Page 44 of 98


(weights and bias). Therefore, when the network model is chose, the performance of the NN configuration
depend on the choice of the number of neuron of layers and the training. As a result of we just got the data of
satellite images and gauge, several configurations of feed-forward networks have been trained only selected
satellite images form the phases sequence of 1 and 7 June 2007. A further improvement in the performance of a
certain neural configuration is expected next we will use wider set of input patters, representative of different
meteorological situations is provided to the network in the training phase. In addition, the configuration of feed-
forward network will be done lots of experiments, choosing the best as we need.

Table 5.1. Input variables for TLFNN model
IPA The ratio of the intensity IR3 and the sum of brightness temperature IR1
(IR3/(TBBIR1+TBBIR2)) IR2
INCREMENT The infrared brightness temperature IR1 increment of the pixel hourly
adjacent interval
(TBBIR1) The infrared brightness temperature of the pixel
The grad of brightness temperature of 3#3pixel window centered on the
target pixel
LAT The latitude of infrared brightness temperature IR1 of the pixel
LONG The longitude of infrared brightness temperature IR1 of the pixel
The mean of brightness temperature of the 3#3pixel window centered on
the target pixel
The standard deviation of brightness temperature of the 3#3pixel window
centered on the target pixel
The mean of brightness temperature of the 5#5pixel window centered on
the target pixel
The standard deviation of brightness temperature of the 5#5pixel window
centered on the target pixel

As figure 5.1 and table 5.2 below show, the accuracy in this experiment is not high, reasons may be among
these: 1) a too short phase used for training (1 and 7 June 2007); 2) the number of neuron of layers is not the
best; 3) the choice of phase for training is not typical; 4) the performance of TLFNN should be compared with
more statistical data with gauge; 5) other reasons to be investigated in the next months.

Page 45 of 98


Fig 5.1.The difference of Gauge and TLFNN on 8/26,2007(00:00~23:00)

Table 5.2. Comparing of Samples statistics of Gauge and TLFNN on on 8/26,2007(15:00)
Rainfall[mm] Gauge TLFNN
No rain 31 25
0~2 14 12
2~10 5 14
>10 2 1

References
Sugimoto S., K. Ueno and W. Sha, 2008: Transportation of water vapor into the Tibetan Plateau in the case of a
passing synoptic-scale trough. J. Meteor. Soc. Japan, 86, 935-949.
Ueno K., S. Takano and H. Kusaka, 2009: Nighttime precipitation induced by a synoptic-scale convergence in
the central Tibetan Plateau. J. Meteor. Soc. Japan, 87, 459-472.

Page 46 of 98


Summary of progress

WP6 aims at the estimation of glaciers and snow melt-water on the QTP. During the period from May 1st 2008
to October 1st 2009, works progressed toward the objectives and scheduled tasks as planned. In the first level,
the algorithms for snow/ice and frozen soil properties retrievals and products, that major on snow cover/fraction
cover, snow water equivalent(SWE), soil freeze/thaw status, have been reviewed and intercompared. With the
above works as base, the prototypes for mapping snow cover/fraction cover, SWE retrieval and soil freeze/thaw
status classification are primarily developed and their daily products has been generated and validated.

In addition, we are developing a high-resolution meteorological dataset, which will be used to drive the land data
assimilation system and derive soil parameters. Moreover, the mass balance observation for Zhadang glacier and
hydrological as well as meteorological observations in this region have been carried out as the first step to
evaluate glacier area and volume on QTP. Finally, climate effects on glacier mass balance was modelled using
“positive degree day factor” method and a new snowmelt model (Binggou Snowmelt Model (BSM)) which can
couple remote sensing data and meteorological observation was established using snow energy balance method.

Task 6.1: Review and intercomparison of available algorithms for snow/ice properties (fraction cover,
water equivalent) retrievals and products (CAREERI, BNU, TUD)

(1) A review has been performed of the available algorithms for snow/ice properties. The focused
algorithms are the ones for snow cover/fraction cover mapping, SWE retrievals by passive microwave remote
sensing and soil frozen/thaw status identification.

(2) Satellite datasets both of the optical, such as the MODIS and microwave like SMMR and SSM/I have
been collected. Besides, the global products of snow cover and SWE on QTP has been validated by some in situ
observations now available to us and by the results of high resolution images like TM. Due to the forgone work
and the want of more general assessment, some validation on snow cover mapping products has been conducted
in Xinjiang Province as for reference. The drawbacks of NSIDC global snow cover products and the necessary
of developing a new algorithm for retrieving snow depth on QTP have been identified.

The Dept. of Remote Sensing of Delft UT contributes to two tasks within WP 6. For both tasks, data of the
GLAS full waveform laser ranger on board of the ICESat mission has been used. GLAS collected world wide
along-track elevations between 2003 and 2009 with a decimeter vertical accuracy. It has no side-looking
possibilities, therefore sampling over Tibet is strongly hampered by its near polar orbit, resulting in an across
track distance between adjacent tracks of 70 km. ICESat records the full waveform return signal, which means
that the signal return resulting from the convolution of the outgoing signal with the vertical structure in the ~70
m footprint is sampled at 15 cm vertical resolution. From this return signal an elevation is estimated using a
uniform method.
Ongoing work links the shape of the GLAS full waveform returns from GLAS data over the Nyainqentanglha
Mountains to land cover and glacier characteristics. It is considered to what extend the full return signal can be
directly used to decided on properties of the surface. In a first study, a distinction is made between returns from
water (Namtso lake), rock and glacier. In a second study, variations in return signal within a glacier (Bare ice,
snow, debris) will be considered.

Using ICESat repeated laser range data it is possible to analyze changes in elevation along track during the
mission lifetime from 2003 and 2009. However, tracks are only repeated up to a few 100 m. Therefore direct
monitoring of glacial elevation changes is challenging. It has been demonstrated however, (Figure 6.2), that
ICESat is very suited to obtain elevation changes over many Tibetan lakes. In a next step the links between

Page 47 of 98


glaciers (CAREERI glacier mask) and lakes (MODIS water mask) will be established using GIS techniques in
order to, at least partly, relate lake elevation changes to meltwater equivalents.

Task 6.2: Developing a prototype for mapping snow cover extent and Snow Water Equivalent (SWE)
(CAREERI, BNU)

(1) We have developed a new daily snow cover mapping algorithm by: 1) improving the NSIDC snow
covering algorithms and 2) combining MODIS-Terra and MODIS-Aqua data on QTP. Further more, a method
for mapping fractional snow cover from MODIS data on QTP has also been proposed. The new snow cover
mapping algorithm can provide daily snow cover products at 500-m resolution on QTP. The new snow cover
algorithm employs the CIVCO topographic correction, a grouped-criteria technique using the Normalized
Difference Snow Index (NDSI) and other spectral threshold tests and image fusion techniques to identify and
classify snow on a pixel-by-pixel basis.

(2) We Modified the Chang snow algorithm to make it suitable for snow depth retrieval on QTP using
SMMR and SSM/I remote sensing data and snow depth data recorded at the China national meteorological
stations. We further analyzed the accuracy and uncertainty of the new snow product generated by the modified
Chang algorithm. The daily snow depth datasets in China from 1978/1979 to 2005/2006 has been produced, and
their spatial and temporal characteristics analyzed primarily.

(3) We have developed a new decision tree algorithm to classify the surface soil freeze/thaw states. The
algorithm uses SSM/I brightness temperatures recorded in the early morning. Three critical indices are
introduced as classification criteria—the scattering index (SI), the 37 GHz vertical polarization brightness
temperature (T37V), and the 19 GHz polarization difference (PD19). And the discrimination of the desert and
precipitation from frozen soil is considered, which improve the classification accuracy. Long time series of
surface soil freeze/thaw statuses can be obtained using this decision tree, which potentially can provide a basic
dataset for research on climate and cryosphere interactions, carbon cycles, hydrological processes, and general
circulation models

Task 6.3: Evaluating glacier area and volume on the Qinghai-Tibet Plateau (CAREERI, BNU, ITP, TUD).

(1) Observation of mass balance for the Zhadang glacier;
(2) Hydrological and meteorological observation in the Zhadang glacier area
(3) Modelling climate effects on glacier mass balance using “positive degree day factor” method
As it is relatively easy to estimate changes in the glacier area using readily available image data, Delft UT has
focused on novel methods aiming at analyzing volume changes using different types of topographic satellite
data. First SAR data can be applied in two essentially different ways to obtain glacial flow velocity fields. One
method is image based, and obtains velocities by matching images obtained from different moments. This
method has been successfully applied by others on e.g. Baltoro. The other method, InSAR, exploits phase
differences between different acquisitions. This method has been used to obtain a flow velocity map of Rongbuk
glacier, on the North side of Everest.

Topographic coverage of all Tibetan glaciers could be obtained using photogrammetric techniques applied on
stereo data from e.g. the ALOS/PRISM instrument. Efforts are ongoing to optimally profit from the information
contents in the ALOS/PRISM imagery. First results at the border area between China, India and Nepal show
DTM processing using standard software is possible, but is hampered by low texture on snow, and shadow
effects and low visibility in deep valleys. A next step will consider custom made software to overcome part of
these problems.

Page 48 of 98


Task 6.4: Providing soil parameter data sets for the entire plateau from a microwave land data
assimilation system from 2008 to 2010

(1) We are developing a high-resolution meteorological dataset, which will be used to drive the land data
assimilation system and derive soil parameters. The forcing data includes six components: shortwave radiation,
longwave radiation, precipitation, pressure, wind, air temperature, and air humidity. Several data sources are
used for the development: in situ data from China Meteorological Administration (CMA), precipitation from
GPCP, and other components from Princeton University. At this moment, we are focusing on the shortwave
radiation. Preliminary results shows the accuracy of radiation can be significantly improved compared with
currently available forcing data.
(2) To facility the land data assimilation system, we have taken a number of soil samples, with a total
weight of 80 kg. The soil samples are being tested at Institute of Tibetan Plateau Research to measure its
hydraulic and thermal parameters. The output of laboratory experiments would provide a basis to evaluate land
surface models and the data assimilation system.

Task 6.5: Estimation of glaciers and snow meltwater. (CAREERI, BNU):

Snow energy balance method was utilized to set up a new snow model - Binggou Snowmelt Model (BSM) - with
remote sensing data and meteorological observation. Snow distribution, snowmelt and snow sublimation were
modelled by BSM in 2008 snow season in Binggou watershed.

Significant results

1. On Snow Cover Mapping and SWE Retrieval:
We proposed a modified algorithm to retrieve the snow depth on QTP from SMMR (1978 to 1987) and SSM/I
(1987 to 2006), and analyzed the spatial and temporal variations of snow depth over whole QTP. The snow
depth products were generated based on the new algorithm from SMMR and SSM/I during the period from 1978
to 2006 on the whole QTP.

We have developed a new daily snow cover mapping algorithm based on which daily snow cover products at
500-m resolution on QTP has been generated. In addition, a prototype for snow fraction cover mapping has been
proposed.

A decision tree algorithm was developed to identify the surface soil freeze/thaw states by taking the influence of
the desert and precipitation into account. The more reliable SI was introduced into this decision tree instead of
SG to identify the scatterers. The average accuracy of the classification result was 87%, which was validated
against the 4 cm deep soil temperature observations. Most misclassifications occurred when the soil
temperatures were near the soil freezing point and during the transition period between the warm and cold
seasons. A grid-to-grid Kappa analysis was also conducted to evaluate the consistency between the map of the
actual number of frozen days obtained using the decision tree classification algorithm and the map of
geocryological regionalization and classification in China. The results showed that the overall classification
accuracy was 91.7%, while the Kappa index was 80.5%. Both validation results show that this new decision tree
algorithm based on SSM/I brightness temperature can produce a long time series of surface soil freeze/thaw
status from the launch of SSM/I in 1987 until now with an accuracy capable of providing a dataset to analyze the
timing, duration and areal extent of surface soil freeze/thaw status for the research on climate and cryosphere
interactions, carbon cycles, and hydrological processes in cold regions.

Page 49 of 98


Fig. 6.1 Actual number of frozen days in China for the period from Oct. 1, 2002 to Sep. 31, 2003

2. On Glacier Area and Volume Evaluation
Observations in the Zhadang glacier indicate the parameterization of mass balance with annual precipitation
amount is insufficient to describe the response of glaciers to climate change. Seasonal concentrations of
precipitation strongly influence glacier mass balance, especially in monsoon regions with summer precipitation
climate.

Using mass balance and meteorological data in the ablation season of the year 2007 and 2008 in Zhadang
glacier, degree-day factors have been obtained for snow (5.3 mm•d-1• -1) and ice (4.0~14.0 mm•d-1• -1 at
different altitudes with an average of 9.2 mm•d-1 -1). Degree-day factors for the Zhadang glacier drop with
elevated altitude, though there are no significant changes along with time. Mass balance in 2006/2007 and
2007/2008 of Zhadang Glacier is estimated using degree-day model. The simulated mass balance in 2006/2007
is -393.2mm w.e., and 289.7mm w.e. for the year 2007/2008.

3. On Soil Parameters Dataset for Whole QTP from Microwave Assimilation System
We have accomplished the validation of the land data assimilation system, and this work has been published by
Journal of Hydrometeorology, Vol. 10 (3) (Yang et al., 2009; see the attachment). The auspice of CEOP-AEGIS
is clearly acknowledged. This study testifies the capability of a new microwave land data assimilation system
(LDAS) for estimating soil moisture in semi-arid regions, where soil moisture is very heterogeneous. This
system assimilates the AMSR-E 6.9 GHz and 18.7 GHz brightness temperatures into a land surface model

Page 50 of 98


(LSM), with a radiative transfer model as an observation operator. In order to reduce errors caused by
uncertainties of system parameters, the LDAS uses a dual-pass assimilation algorithm, with a calibration pass to
estimate major model parameters from satellite data and an assimilation pass to estimate the near-surface soil
moisture.

4. Validation data of soil moisture were collected in a Mongolian semi-arid region. Results show that (1)
the LDAS-estimated soil moistures are comparable to area averages of in situ measurements, though the
measured soil moistures were highly variable from site to site; (2) the LSM-simulated soil moistures show less
biases when the LSM uses LDAS-calibrated parameter values instead of default parameter values, indicating that
the satellite-based calibration does contribute to soil moisture estimations; (3) compared to the LSM, the LDAS
produces more robust and reliable soil moisture when forcing data become worse; the lower sensitivity of the
LDAS output to precipitation is particularly encouraging for applying this system to regions where precipitation
data are prone to errors.

Figure 6.2 Tibetan lake level trends between 2003 and 2009 as captured from ICESat data

5. On Glacier and Snowmelt Water Estimation
Binggou Snowmelt Model (BSM) was designed to model snow water equivalent and runoff, and the results were
validated by measured snow depth. Daily snow cover observation was simulated by a utilization of daily and
eight-day MODIS snow cover products and meteorological observations. With the simulated SWE, snowmelt
processes at both the point-scale and basin scale were analyzed in detail. The net energy input was negative
before snowmelt occurrence and heat eradiated from snowpack in this period. With solar azimuth changed,
shortwave radiation enhanced and air temperature increased, energy input into snowpack increased and resulted
as three large-scale snowmelt processes. Meteorological measurement, field observation and daily runoff data
were used to validate the simulation results by BSM. The results were in agreement well with three different
observations but with some problems because: 1) the point-scale measurement could not be represented by grid
simulation; 2) snow cover was not recognized well sometimes and 3) frozen and thaw soil was not considered
properly.

6. Glacier flow and mass balance

Page 51 of 98


The analysys of ICESAT, ALOS/PRISM and ENVISAT/ ASAR data led to the following results:
$ Lake level changes of over 100 Tibetan lakes between 2003 and 2009 (Fig.6.2)
$ Subdivison of lake level changes into major river basins
$ Determination of corresponding changes in water volume
$ Flow velocity map of Rongbuk glacier
$ Construction of ALOS/PRISM DTM using standard software (Border area China/India/Nepal) using
ICESat ground control points
$ Validation ALOS/PRISM DTM using ASTER-GDEM and independent ICESat elevations.

Page 52 of 98


Summary of progress

Numerical Weather and Climate Prediction modeling system”
K. Ueno (University of Tsukuba) and X. Shen (Chinese Academy of Meteorological Sciences )

3.7.1 Progress for detailed analysis of the relationship between Plateau land surface processes, monsoon
onset and intense precipitation with a coupled land-atmosphere meso-scale model by UNITSU
In the WP7, University of Tsukuba (UNITSU) team is mainly focused on the 1nd objectives “Detailed analysis
of the relationship between Plateau land surface processes, monsoon onset and intense precipitations with a
coupled land – atmosphere meso-scale model”. Based on the analysis results, UNITSU is also responsible for
nominating candidate cases of precipitation systems for CAMS team to assess the land-surface effects and
improving lead time in the forecasts of intense precipitations by the numerical sensitivity studies. Then, our team
followed three steps in the analysis during 2008-2009.
First, a target period was set through January to September in 2008 which covered seasonal transition of land-
surface condition in spring and monsoon onset periods. The year of 2008 corresponded to “Asian Monsoon
Year”, and JICA intensive observations were also conducted, that provided good opportunity to archive multiple
in-situ data sets to validate products by WP7. WP7 is basically planning to use land-surface data produced by
other working packages. However, the products have not been ready at this moment. Then, UNITSU archived
existing global data sets, such as satellite and re-analysis data products by agencies as listed below, and have
started assessments;
1) MODIS/Terra Snow Cover 8-day L3 Global 500m Grid, Ver. 5 (snow cover)
Jan.-Sep 2008, 8-day composited , 500m 500m
National Snow and Ice Data Center ;NSIDC (http://nsidc.org)
2) AMSR-E Level 3 Daily Soil moisture, Ver. 6 (Soil moisture)
Jan.-Sep. 2008, Daily, 0.25 *0.25
Japan Aerospace Exploration Agency; JAXA (http://www.eoc.jaxa.jp/iss/index.html)
3) JRA25 ( Dec. 2004)+JCDAS (Jan. 2005 ) (Geopotential height, Air temperature, Specific humidity, Dew
point depression, Zonal wind, meridional wind, Cloud water content, Sea level pressure)
Jan. 1990- Dec. 2008, 6-hourly, 1.25 *1.25
Japan Meteorological Agency- Central Research Institute of Electric Power Industry
(http://jra.kishou.go.jp/)
l JCDAS takes over the same system as JRA25, and the data assimilation cycle is extended up to the
present (More detail information is shown in a web site of JRA25 ).
4) Global Precipitation analysis; GPCP (precipitation; mm/day)
Jan. 1997-Apr. 2008 (May-Sep. 2008 unreleased yet), Daily, 1 *1
NASA/Goddard Space Flight Center; GSFC (http://precip.gsfc.nasa.gov/)
5) METEOSAT7 (IR; This satellite has only 10.5-12.5 µm )
Jan.-Sep. 2008, Hourly, inhomogeneous distributed data (about 5 km at sub-satellite point)
EUMETSAT (http://www.eumetsat.int)
6) FY2 (IR1=10.3-11.3µm, IR2=11.5-12.5µm, IR3=6.3-7.0µm, IR4=3.5-4.0µm)
Jan.-Sep. 2008, Hourly, 0.04 *0.04 , 44.6E 164.6E,60S 60N
Center for Environmental Remote Sensing (CEReS) 4 Virtual Laboratory (http://www.cr.chiba-
u.jp/~4vl/wiki/wiki.cgi)
7) Global Surface Summary of Day; GSOD (observation data at ground surface)
Jan.-Dec. 2008, daily (only in China)
National Climatic Data Center (NCDC); NOAA
(http://www.ncdc.noaa.gov/cgi-bin/res40.pl?page=gsod.html)

Page 53 of 98


Secondary, seasonal transition was examined by JRA25 data and meso-scale convective systems (MCS) were
automatically identified using the METEOSAT data followed by the method of Evans and Shemo (1996).
Global data sets showed that there were several periods for plateau-scale snow cover during winter and normal
onset of monsoon in the middle of June. Then, we focused on the strong convective activities in the warm
season, and extracted the MCS occurrence and movement automatically. Occurrence of the MCS was
coincident with intra-seasonal variability in the mid-latitudes, and dominated in the two areas over the TP, such
as south and southeast. In the periphery of the TP, MCS tended to occur along the south of Himalayas, Assam,
and over the extending mountain zone from the southeastern TP, and they were mostly activated during the
night. The MCS was divided into two groups according to the synoptic conditions, such as 1) strong surface
heating conditions with prevailing of upper high-pressure system (Tibetan High) without effects of mid-latitudes
disturbances or troughs, and the other is 2) the severe weather cases in the lee-side of TP associated with troughs
or apparent fronts without the Tibetan High. Candidate periods for each condition are listed as follows, and
handled to CAMS team;

Thirdly, we examined the influence of TP for development of MCS by numerical simulations for some cases of
two types. Numerical simulation in the UNITSUK was designed by Weather and Research Forecast (WRF)
model with three nesting domains, and non parameterization with two-way interactive simulation was conducted
with 4 km resolution in the last domain. Occurrence of MCSs in the model was well corresponded with
METEOSAT images, and processes of development was diagnosed. Tentative results were introduced at the
joint international conferences of IAMAS/IAPSO/IACS in Montréal, Canada by Sugimoto and Ueno (2009) and
Ueno (2009).

References
Shiori SUGIMOTO and Kenichi UENO, 2009: The effect of synoptic and land-surface conditions for
precipitation processes over the Tibetan Plateau, MOCA-09, the IAMAS/IAPSO/IACS 2009 Joint
Assembly, Abstract No. 402, July, Montreal, Canada.
Kenichi UENO, 2009: Mountain weather modification in the Tibet/Himalayas, MOCA-09, the
IAMAS/IAPSO/IACS 2009 Joint Assembly, Abstract No. 514, July, Montreal, Canada.

3.7.2 Progress for improving lead time in the forecasts of intense precipitations by CAMS
2.1 Brief introduction of GRAPES_Meso
The GRAPES is a unified NWP model with 3/4DVAR data assimilation system, which is the abbreviation of
Global/Regional Assimilation and PrEdiction System. The main features of GRAPES include: (1) fully
compressible equations with hydrostatic/non-hydrostatic option; (2) the semi-implicit and semi-Lagrangian time-
stepping method; (3) height terrain-following coordinate in the vertical and latitude-longitude spherical
coordinate in the horizontal; (4) scalar advection by piece-wise rational method; (5) fully physical package.
The meso-scale version of GRAPES is utilized in WP-7. Its main characteristics are listed up in the following
table.
Flux-form equations of water substances

Piece-wise rational method + volume-

Page 54 of 98


remapping SL for scalar advection
Dynamics
Reference atmosphere based on the initial
field
Effective topography
NOAH LSM + Simple initialization
Xu and Randall Diagnostic cloud
Betts-Miller-Janjic Cumulus
CAMS mixed-phase microphysics
Physics Effect of slope on surface radiation

Significant results

Re-run of GRAPES_Meso with 15km horizontal resolution for 2008 has been finished, and verification of
rainfall forecast over Tibet was conducted. Figure 1 gives the time series of forecasted and observed 24hr
accumulated rainfall. As shown, GRAPES_Meso can well capture most of the rainfall events occurred in Tibet
during the period from April to September of 2008. This encourages us to further investigate the possible role of
underlying surface-atmosphere interaction on convective activities in the future work.

averaged area 70~105 E 22.34~40
N

Fig.7.1: Time series of forecasted and observed 24hr accumulated rainfall averaged over area (70~105 E
22.34~40 ). Unit is mm.

(2) Through case study, investigation of effect of different complexity LSM on convective initiation has been
conducted. In the study, SLAB and NOAH (developed by Oregon State University) land surface models (LSMs)
were employed to understand the impacts of different land surface processes on the initiation of convective
activities. A locally-developed convection case occurred on August 2, 2003 in Jiangxi Province of China was
selected. Figure 7.2 shows the simulated (fig. 7.2.b-e) and observed rainfall (fig. 7.2.a). Clearly, the simulated
rainfall by using complex NOAH LSM exhibits closer to the observations.

Page 55 of 98


(a)

(b) (c)

(d) (e)

Fig.7.2: (a) observed 24-h precipitation, (b) simulated 6-h precipitation from 00 to 06 UTC 02 Aug 2003 in color scheme by
NOAH and by SLAB (c), simulated 12-h precipitation (00-12UTC) by NOAH (d), and by SLAB (e). Unit: mm. The box
shows the observed precipitation band over north-east Jiangxi.

Further analysis show that regional convective precipitation is extremely sensitive to the land surface
processes. The NOAH model applied in the study had a rational simulation of the initiation of convective
activities, while the SLAB model produced a retarded initiation of convective activities by 1-2 hours, implying
that NOAH is good at describing surface sensible and latent heat. Soil temperature and moisture have a direct
impact on the distribution of surface sensible and latent heat. Distribution of surface sensible heat flux, in turn,
affects the development of boundary layer. The development of boundary layer affects the onset of local
circulations, by altering the stability of thermal-dynamic structures of the boundary layer, and by directly
affecting the initiation of convective activities. NOAH made a quick response to the increased sensible heat flux

Page 56 of 98


form nighttime to daytime. Booming sensible heat flux facilitated the fast development of the boundary layer,
with correspondingly enhanced convective available potential energy (CAPE), creating the needed conditions for
kicking off convective activities. A rational and detailed land surface process is extremely important for
numerical modeling, especially for the initiation and development of a strong local convective system under the
weak synoptic forcing.
3 Systematic evaluation of effects of LSM on daily rainfall forecast for a successive heavy rainfall event was
conducted. The successive heavy rainfall has been occurred over the Huai River basin during the period from
end of June to July 11 in 2007. The successive 24-hour rainfall forecast by using GRAPES_Meso with one-way
nested 16km 6km and 2km resolutions shows the obvious improvements were found in simulations of location
and intensity of heavy rainfall by using NOAH LSM. The threat score (TS) of precipitation becomes larger than
that by using the simple SLAB.

Page 57 of 98


Summary of progress

Task 8.1: Evaluation of water balance calculation approaches
FutureWater, ArieSpace and IGSNRR have assessed a number of water balance calculation approaches on
their usability as water balance monitoring systems. Several criteria were used to rank the different
approaches:

• The spatial land surface schematisation of the model should be detailed in such a manner that
integration with E.O. products is warranted.
• The model should recognize direct forcing of daily grids of ET, precipitation and snow cover
• Model outputs include all components on the water balance at sufficient level of temporal and spatial
detail. This is specifically true for top soil moisture, which will be used for validation.
• The model includes algorithms related to snow melt, infiltration, surface runoff, drainage, percolation
and groundwater base flow.
• The model should be able to incorporate glacial melt in a lumped mode.
• The model should be able to deal with permafrost processes.
• The model needs to include a (horizontal) routing component that allows accumulation of runoff
across the Tibetan plateau.
• The Tibetan plateau includes number of large lakes and artificial reservoirs. The model routing
component needs to be able to take into account storage variation in lakes and artificial reservoirs and
the resulting delay in water yield.
• The model source code must be accessible and customizable.

Based on these criteria a number of candidate models were evaluated. These models include SWAT, HBV,
LARSIM, GRAPES and PCRGLOB-WB. The analysis showed that all the models, except SWAT and
GRAPES, have similar algorithms for soil water balance and runoff. SWAT is based on a different method
and it has probably a slightly more advanced routing, based on Muskingum approach and GRAPES has a
more advanced soil water balance module but no routing. Most model require source code changes to allow
forcing by remote sensing (precipitation, evapotranspiration, snow), except for PCRGLOB-WB which
already has an inbuilt option for pre-described actual evapotranspiration. Some changes are still required to
include snow cover forcing. GRAPES, LARSIM and PCRGLOB-WB are raster based, but GRAPES has no
routing component. LARSIM and PCRGLOB-WB are efficient, open source and with very close links to the
original developers which allows straightforward customizing in this project context. Based on this analysis it
was decided that the PCRGLOB-WB is used as water balance monitoring tool for the entire plateau, but that
at local scale other models are tested as well. In particular the HIMS model that is developed at IGSNRR.
This model will be developed simultaneously with the plateau model for the upper Yellow river catchment
and can be used to validate the plateau model. The results of the evaluation of the water balance approaches
are reported in (1).

Task 8.2: Water balance and run-off calculation over the entire Plateau

Page 58 of 98


FutureWater has setup a first concept version of the water balance monitoring system of the Tibetan plateau
based on the PCRGLOB-WB model. The model setup at a high spatial resolution (5km) and simulated the
complete hydrological cycle on a daily step. In each cell the vertical flow of water through four compartments
(canopy and three soil compartments), soil and canopy are fed by rainfall and snowmelt and depleted by
evapotranspiration. Runoff and groundwater base flow are transferred to the to the drainage network and
routed along the digital elevation model. Discharge is calculated from the kinematic wave approximation of
the Saint-Venant equation. A schematic overview of the plateau model including the data requirements is
shown in.

Figure 8.3 Schematic overview of the Tibetan plateau water balance model

Currently the model is forced by public domain reanalysis data based of the ERA40 dataset
(evapotranspirtation and temperature) and on TRMM data (precipitation). As the project progresses the
model will be forced and validated with datasets from the other WPs. For the coming period a number of
conceptual improvements and additions are planned:

• Evaluate and possibly modify soil water algorithm (e.g. compare with HIMS)
• Incorporate reservoirs in routing scheme
• Incorporate model for glacier melt
• Forcing with data from other CEOP-AEGIS WPs
• Further detail soil and vegetation parameterization based on RS datasets
• Validation with soil moisture

Page 59 of 98


Task 8.3 : Water balance and climate change
IGSNRR has collected geographic information (Fig.8.2) including DEM (digital elevation model), river
network, land use and land cover, soil type of the entire plateau has been prepared for running the water
balance model. Long-term climate and hydrologic data of the Qing-Tibet Plateau has also been collected for
the period 1960 to 2000 from 89 stations (Fig.8.3). Spatial interpolation has been processing to provide 10
10km climate dataset for the entire plateau considering the location and the elevation of the grids. The sample
dataset of the Headwater of the Yellow River Basin is around 1.33GB. The details of the dataset are as
followed:
(1) Spatial Resolution: 0.1 degree, 1254 grids;
(2) Temporal Resolution: Daily;
(3) Periods:1960 2001;
(4) Climate variables:
" Mean Daily Temperature: oC
" Maximum Daily Temperature: oC
" Minimum Daily Temperature: oC
" Vapor Pressure: hPa
" Air Pressure: hPa
" Wind Speed at 2m: m/s
" Sunshine Duration: hours

Figure 8.2 DEM and land use/cover of the Qing-Tibet Plateau

Page 60 of 98


Figure 8.3 Climate stations of the entire plateau and the sample dataset of the Headwaters of the Yellow
River Basin

Significant results

According to the climate data collected, the long-term climate and hydrologic changes of the plateau have
been detected using the Mann-Kendall method. The results have shown that potential evapotranspiration,
wind speed and solar radiation tends to decrease during the past 40 years, while temperature and vapour
pressure deficit tends to increase. The increasing rate of temperature was about 0.28oC/10a (Fig.8.4). The
spatial patterns of the change showed that temperature (Tmax, Tmin, Tmean), relative humidity (RH) and
precipitation (P) increased in most part of the plateau, while potential evapotranspiration (ET0 and ETpan),
wind speed (U) and sunshine duration (Shour) decreased. The vapour pressure deficit (VPD) increased in the
north part of the plateau while decreased in the south part (Fig.8.5).

Page 61 of 98

(0C/y)
0.028SyearTmean
(kPa/y)197019751980198519901995200022.533.544.5slope:
(m/s/y)19701975198019851990199520000.40.420.440.460.480.5slope:0.00028NSyearVPD
0.017SyearU2
(MJ/m2/y)19701975198019851990199520001.522.53slope:
1.43SyearRn
(mm/y)19701975198019851990199520003200325033003350slope:
4.57SyearETpan
(mm/y)1970197519801985199019952000150016001700180019002000slope:
1.05SyearET0
(0C/y)197019751980198519901995200092094096098010001020slope:
0.028SyearTmean
(kPa/y)197019751980198519901995200022.533.544.5slope:
(m/s/y)19701975198019851990199520000.40.420.440.460.480.5slope:0.00028NSyearVPD
0.017SyearU2
(MJ/m2/y)19701975198019851990199520001.522.53slope:
1.43SyearRn
(mm/y)19701975198019851990199520003200325033003350slope:
4.57SyearETpan
(mm/y)1970197519801985199019952000150016001700180019002000slope:
1.05SyearET0
197019751980198519901995200092094096098010001020slope:
CEOP-AEGIS (GA n° 212921)

Page 62 of 98
Figure 8.4 Long-term climate changes of the Qing-Tibet Plateau
Periodic Report no. 1


Figure 8.5 Spatial patterns of long-term climate change in the entire plateau (1960-2001)

To assess the impacts of climate and land surface change on streamflow, an approach based on the concept
climate elasticity has been proposed assuming that:
(8.1)

Page 63 of 98


where Q is streamflow, P and E0 are precipitation and potential evapotranspiration respectively representing
dominant climate factors on hydrological cycle, V is a factor that represents the integrated effects of catchment
characteristics on streamflow. Following Eq.8.1, changes in streamflow due to changing climate and catchment
characteristics can be approximated as:

(8.2)
where , , and are changes in streamflow, precipitation, potential evapotranspiration, and
catchment characteristics respectively, with , and . On the

assumption that the land surface factors are independent of the climate factors, Eq.8.2 can be rearranged as:
(3a)
(3b)
(3c)
where , are changes in streamflow due to climate change and land use/cover change respectively. In
Eq.8.3a, can be estimated from observed streamflow records, thus if or is known, the
framework can be used to separate the effect of climate change from that of land use/cover change on
streamflow. The effect of climate change on streamflow ( ) was assessed using the concept of climate
elasticity of streamflow defined as:

(4)

Thus, Eq.8.3b can be rewritten as:
(5)

where and are elasticity of streamflow with respect to precipitation and potential evapotranspiration.
The case study on the Headwaters of the Yellow River Basin (HYRB) have shown that land use change is
responsible for about 74.6% of the streamflow reduction in the 1990s, while climate change contributed to
25.4% of the reduction. The climate elasticity appears to have an inverse relationship with runoff coefficient, but
positive relationship with aridity index, showing that the drier the catchment, the more sensitive streamflow is
with respect to precipitation change.

References
1. Immerzeel, W. et al., Model selection for the Tibetan plateau water balance monitoring system (CEOP
AEGIS report, Strassbourg, 2009), pp. 1-59.
2. Zheng, H., L. Zhang, R. Zhu, C. Liu, Y. Sato, and Y. Fukushima (2009), Responses of streamflow to
climate and land surface change in the headwaters of the Yellow River Basin, Water Resour. Res., 45,
W00A19, doi:10.1029/2007WR006665.

Paper Published:
Zheng, H., L. Zhang, R. Zhu, C. Liu, Y. Sato, and Y. Fukushima (2009), Responses of streamflow to climate and

Page 64 of 98


land
surface change in the headwaters of the Yellow River Basin, Water Resour. Res., 45, W00A19,
doi:10.1029/2007WR006665.

Page 65 of 98


Summary of progress

Task 9.1: Identification of study areas and ground data collection both in China and India (Alterra, BNU,
IRSA, NIH).
Information on past drought events, damage on agriculture resulted by drought both in China and India were
collected and reviewed. Table 9.1 gives the summary of severe drought events in the past, while Fig 9.1 is the
summary of drought prone areas in India.

Pilot areas in both countries are preliminarily identified according to above information. The pilot area in China
will be the North Plain – part of Yellow River basin (for instance Henan province) and southwest area – part of
Yongtz river basin (Sichuan-Chongqing ). In India the pilot area will be Ganga River basin.

Historical meteorological data (air temperature and humidity, precipitation etc) were collected and analyzed over
the pilot areas. This ground dataset will also be taken as reference in the development and evaluation of
algorithms for anomaly detection by using satellite data (different land surface parameters) to ensure systematic
analysis on drought events over study areas.

Part of GIS data (for instance shape files of boundaries of administrative areas at country, province and county
levels) have been collected over China. The further GIS data are under collection.

Figure 9.1: (Left) Natural hazards areas in India; (Right) key vulnerable river basins in India.

Page 66 of 98


Task 9.2: Vegetation dynamic monitoring through long-term time series of satellite observations for
China and India (UVEG, NIH, IRSA)
A bibliographical review of methods for monitoring vegetation has been carried out. The climate and vegetation
changes undergone by the Tibetan plateau during the past decades have been identified. Pathfinder AVHRR data
have been downloaded and checked for consistency. These data have been resized to the study area. Algorithms
have been identified and implemented in IDL language in order to obtain NDVI (Normalized Difference and
Vegetation Index) and LST (Land Surface Temperature) parameters. The whole data base is presently being
processed at the Global Change Unit of the University of Valencia.

Deviation:
- Analysing the linkage of spatial and temporal vegetation dynamics with changes in climate systems and water
resources on the Tibetan Plateau and surrounding areas over a long-term period (UVEG).
- correlation between drought on the Plateau atmospheric circulation and precipitation on Tibetan Plateau and
surrounding areas. (UVEG/NIH/IRSA).
These two sub-tasks will be done together with task 9.3 once the analysis on rainfall anomalies are completed.

Page 67 of 98


Table 9.1 Summary of severe drought events in the past over China.
Year Region Type Duration Drought Damage
southern areas of Yunnan, the northeast An area of 4.349 million hectares of crops were affected, with an area of
Autumn, winter
regions of Guizhou, the eastern and 0.94 million hectares of crop failure. 51.04 million people were affected,
2010 and spring 2009.10-2010.04
southern parts of Guangxi, Sichuan and 16.09 million people were facing water shortage. A direct economic loss
drought
Chongqing of 19.02 billion yuan
An area of 3.635 million hectares of crops were affected, with an area of
South China (Hunan, Jiangxi,
0.472 million hectares of crop failure. 35.93 million people were
2007 Guangdong, Guangxi, Guizhou and Autumn drought 2007.09 -2007.12
affected, 5.796 million people were facing water shortage. A direct
Fujian)
economic loss of 8.52 billion yuan
An area of 3.776 million hectares of crops were affected, with an area of
0.686 million hectares of crop failure. 66.47 million people were
2006 Chongqing and Sichuan Summer drought 2006.06-2006.08
affected, 15.37 million people were facing water shortage. A direct
economic loss of 22.27 billion yuan
Province-wide drought, serious drought for the past
Heilongjiang
40 years.
Drought area of the crop field is more than 1.53
Jilin
million hectares.
Middle and western of the major grain producing
areas were affected by serious drought, and an area
Northeast China (Heilongjiang, Jilin, Liaoning
2003 Spring drought 2003.2-2003.5 of 2.454 million hectares dry land crop field was
Liaoning and Inner Mongolia)
affected.
Soil moisture of crop field is poor, lots of turning
Inner Mongolia green period pasture dead, cattle weights had a
dramatic decline.
An area of 6.6 million hectares of crops were
Total
affected
North China (Inner Mongolia, Hebei,
An area of 48.76 million hectares of crops and grassland were affected,
2000 Shanxi, Hehan, Gansu, Hubei, Liaoning, Spring drought 2000.02-2000.05
with an area of 5.813 million hectares of crop failure.
Jilin and Heilongjiang)
the area of affected crop field in North China was
June
almost 20 million hectares
Summer and the area of affected crop field in North China was
1997 North China 1997.06-1997.10 July
Autumn Drought almost 26.67 million hectares
an area of more than 6.67 million hectares of planted
October
winter wheat affected
Jiangsu, Anhui, Hubei, Shanghai,
The area of affected crop was up to 30 million hectares, more than 27
Zhejiang, Hehan, Sichuan, Hunan,
1994 Summer drought 1994.06-1994.08 million people and more than 26 million live-stocks were facing water
Jiangxi, Shaanxi, Shanxi, Hebei and
problem.
Shandong
1991 North China (Shaanxi, Hebei and Winter Drought 1990.10-1991.2 An area of 4,840,000 square kilometers was affected

Page 68 of 98


Beijing)
Yellow River and Huaihe River region
(Hubei, Henan, the middle and lower
The area of affected crop field in North China was almost 20.67 million
1988 reaches of Yangtze River, Sichuan, Summer Drought 1988.06-1988.08
hectares.
Guizhou, Liaoning, Shandong, Jilin,
Heilongjiang and Inner Mongolia)
North China (central North China Plain,
The affected area of crop field in Henan province was more than 4
Loess Plateau, Inner Mongolia, Hinggan
1986 Summer drought 1986.06-1986.08 million hectares. In Shanxi province, the affected area was more than 2
League, Henan, Shanxi, Shandong,
million hectares, which accounted for 77% of total planted field.
Hebei, Shaanxi, Sichuan and Gansu)

Page 69 of 98


Significant results
(1) Characteristics of Drought in the pilot area Henan province in China based on SPI analysis
Henan Province is located in the middle of the North China Plain (Fig. 9.2), covering an area of about
167,000 square kilometers which ranges from latitude 31°23!to 36°22! and longitude 110°21! to 116°39!. It
exhibits a transitional climate including the north subtropical humid monsoon climate and the warm
temperate semi-humid climate with average annual precipitation from 600 to 1000 mm, and the altitude
reduces from west to east. Rainfall in this region occurs mainly in summer through the monsoon wind; non-
monsoon rainfall is limited and irregular. Henan province is the largest food producer in China, however, due
to the transitional geographical environment and climatic conditions, precipitation becomes so variable that
drought often occurs and spreads over large areas. Drought is the main natural disaster for agricultural
production in Henan Province.

Figure 9.2 The location of Henan Province, China.

The Standardized Precipitation Index (SPI) was used for the identification of drought events and to
evaluate drought severity in Henan Province, China, in which the SPI is calculated from monthly rainfall
data of 16 meteorological stations from 1952 to 2001. The variation of monthly averaged precipitation in
Henan Province is presented in Fig. 9.3, which shows a typical monsoon climate precipitation pattern, with
rainfall concentration during the summer months, and a very dry winter.

Figure 9.3 Seasonal variation of monthly mean precipitation in Henan Province in China.

Page 70 of 98


The drought severity classes are defined in Table 9.2. In order to estimate the drought frequency in
different seasons, we think that drought event happened if there is at least 1 month of SPI less than -0.5 in a
season. Fig. 9.4 shows the spatial pattern of drought frequency based on SPI value calculated at 1-month
time scale from 1952 to 2001.

Table 9.2 Drought severity based SPI.

Figure 9.4 Spatial pattern of drought frequency in Henan Province based on SPI calculated at the time scale of 1
month.

Influenced by the seasonal movement of West Pacific Subtropical High and Siberian High, Henan Province
is vulnerable to the drought events. The drought frequency is high in spring, summer and autumn. Spring is a
transition season and the inter-annual variability of precipitation is high, which caused the drought frequency
above 50% in the whole province and it is higher in the north. In summer, with the movement of the West
Pacific Subtropical High from the south to north, Henan province gets into the rainy season. However, the
frequently happened abnormity of West Pacific Subtropical High movement brings an extremely large
instability of the rainfall in each month, which caused the high drought frequency in summer, and it is above
60% in most of Henan province. The drought frequency is highest in west area and lowest in east area in
autumn. In winter, the climate in Henan province is controlled by the Siberian High, the precipitation and the

Page 71 of 98


inter-annual variability of rainfall is lowest, the drought frequency decreases and it ranges from less that 20%
in north area to above 60% in south area.
The sensitivity of SPI value to the precipitation changes with time scale of SPI. The shorter time scales
quantify more upper soil water, so there is a great fluctuation when the precipitation changes. The longer
time scales reflect the state of subsoil moisture, surface and subsurface water resources, only long period of
rainfall abnormality can make SPI begin to fluctuate, which is reasonable for drought monitoring, especially
long term drought. Therefore, we also choose a long time scale (12 months) SPI to analyze the temporal
variations of drought in Henan Province, and more detailed drought information is gotten from short time
scale SPI.
Fig. 9.5 shows the temporal change of the SPI calculated at different time scale in Zhengzhou station.
In this figure, the red line indicates the SPI value of -0.5, which is the threshold between drought and no
drought.

(a)

(b)

(c)

(d)
Figure 9.5 Temporal change of SPI calculated at different time scale in Zhengzhou station
(a) 1 month; (b) 3 months; (c) 6 months; (d) 12 months.

Page 72 of 98


Eight drought events are identified in Zhengzhou in the 12-month-scale SPI series (Fig. 9.5(d), Table 9.2),
which are the period 1952-1954, 1959-1962, 1965-1969, 1981-1982, 1986-1989, 1991-1992, 1997-1998 and
2001 respectively. For the 3-month and 1-month scale (Fig. 9.5(a), (b)), the SPI values fluctuate frequently.
From 1952 to 1954, there are two deep SPI valleys, whose values are lower than -2, indicating that extreme
drought occurred (Fig. 9.5(d)). During this period, the short time scaled SPI values also show the conditions
of water loss, although there are obvious fluctuations. SPI values at 1 month scale indicate that there is little
precipitation in summer of 1952, autumn of 1953 and spring of 1954, which led to severe loss of surface
water from 1952 to 1954. The drought continued until the continuous rainfall to supplement the cumulative
loss of surface water in the summer of 1954.

Persistent drought occurred from 1959 to 1962 in Zhengzhou (Fig. 9.5(d)), with maximum intensity in the
summer of 1960. The SPI values at short time scales indicate that almost every month of the precipitation
was less than the average, leading to increasingly heavy accumulated losses of water and reaching its peak at
the summer of 1960.
From 1965 to 1969, another drought event occurred, and the intensity was less than before. What is more,
Zhengzhou experienced wet conditions in some months (Fig. 9.5(d)).
There was almost no long period drought occurred during 1970’s. However, the drought frequency increased
again during 1980’s and 1990’s. The drought events at 1981-1982, 1986-1989, 1991-1992 and 1997-1998
can be observed from Fig9.5(d).

Table 9.2 Long-term droughts from 1952 to 2001 in Zhengzhou, China.
Maximum intensity Time of maximum
Drought period Duration
(SPI) intensity
1952.1 1953.5 17 -2.79 1952.7
1959.7 1961.7 25 -2.66 1960.5
1965.7 1967.6 24 -1.88 1965.9
1968.6 1969.6 13 -2.44 1968.8
1986.7 1987.6 12 -1.78 1987.4
1988.6 1989.4 11 -1.21 1988.11
1991.7 1992.6 12 -1.54 1992.3
1997.7 1998.4 10 -1.69 1998.1

(2) Vegetation dynamic monitoring through long-term time series of satellite observations for China
and India
Within the review of methods for monitoring vegetation, an NDVI based method has been developed
and evaluated, which is described in:

Sobrino, J. A. & Julien, Y. (in press). Global trends in NDVI derived parameters obtained from GIMMS
data, International Journal of Remote Sensing, in press.

This method allows the yearly determination of various NVDI based parameters regarding both
vegetation statistics and phenology, which can then be studied interannually in order to retrieve vegetation
changes.

On a different topic, in order to complete the NDVI and LST time series, which present some gaps due
to atmospheric contamination or instrument failure, a methodology has been developed to interpolate
parameter missing values:

Julien, Y. & Sobrino, J. A. (in press). Comparison of cloud-reconstruction methods for time series of
composite NDVI data, Remote Sensing of Environment, in press.

At the time of the redaction of this report, two thirds of the whole Pathfinder database have been
processed for estimation of NDVI and LST parameters. Figure 1 shows an example of LST for the whole
world and figure 2 for the Tibet area.

Page 73 of 98


Figure 2. Example of global LST map for 21 July 1995.

Figure 3. Example of Tibet LST map for 21 July 1995.

Page 74 of 98


3.10 Work progress in WP10 and achievements during the period
Summary of progress

This workpackage started with some difficulty due to the withdrawal of the National Institute of Hydrology,
India. All tasks, with identical work content and deliverables, were taken over by the National Institute of
Technology, Rourkela by May 1st 2009.. The back log has been recovered, but there has been some obvious
problems with the coordination of the Work Package. This notwithstanding, the most critical objectives have
been achieved, namely the development of an algorithm for time series analysis of a satellite-based indicator
of wetness conditions, potentially useful for flood early warning and the synthesis of available information to
identify flood prone areas in China and India. Work in China has advanced significantly towards
development of models useful to map fllood hazard and study the propagation of floods with the support of
satellite data.

Task 10.1. Flood early warning with wetness indicator derived from low-resolution microwave satellite
data (ULP, BNU)
The theoretical basis of the algorithm to determine open water, flooded land, moist soil and vegetation by
interpreting the difference in horizontal vs, vertical polarization of brightness temperature at 37 GHz, !T37
has been developed. The algorithm makes use of known characteristic values !T37 of open water and bare
dry soil in combination with time series analysis (see De.10.1) to determine flooded and moist area by
inverting a linear mixing model with time-dependent parameters.

In the linear mixing model used by Sippel et al. [1994, 1998], Hamilton et al. [1996], the difference
brightness temperature of vertical and horizontal polarization for each end-member is constant during an
entire year time series. This assumption is possibly correct in tropic zone, because tropic plants do not show
very large seasonal changes. However, for subtropical and temperate plants, the seasonal changes of
vegetation canopy and leaf area index are very large, especially for the cropland. The structure and water
content changing of vegetation is significant in area without flood (Choudhury, 1990). Thus, the linear
mixing model needs to be modified to account for the variation of the vegetation.

AMSR-E on board the Aqua satellite measures radiation at six frequencies in the rage 6.9 – 89 GHz, all dual
polarized, with a constant incident angel of 55˚ since May 2001. 36.5 GHz polarized data from AMSR-E at
local solar time of 1:30 PM is used in this research, of which footprint is 14km by 8km.

A first time series of microwave radiometer data has been constructed using AMSR-E measurements.
The information of 10 major floods in Yangtze River basin since 1980 have been collected, which will be
used to analyst correlation between surface wetness indicator from satellite data and occurrence of flood for
early flood warning.

The case – study summarized under “Significant results”demonstrates the use of microwave data to derive
flood prone areas, as well as flood extent in space and time. The satellite flood identification system used in
this study uses data from the microwave sensors TRMM (rainfall) and AMSR-E (soil moisture). The SRTM
shuttle mission provided digital elevation data.

Task 10.2. Real time flood forecasting using data from the atmospheric-hydrologic network (NIT,
BNU)
A real time flood forecasting model, which is named as XiAnJiang Model, has been developed. In this model
three water sources, including surface runoff, interflow and groundwater runoff, are considered. Considering
the uneven of rainfall and difference of underlaying surface condition in a large basin, runoff is calculated in
sub-basins with same rainfall and underlaying surface. In order to test the modle, it has been used to
forecasting flood of HuaiHe River in China.
A method to forecast floods using hydrological data, remote sensing data and other ancillary data and a
Artificial Neural Network Approach has been developed and evaluated in India. For this purpose Collection
of rainfall data and hydrological data of Kosi basin and in the Gandak river basin both tributaries of the River

Page 75 of 98


Ganga. Maps needed for the numerical simulation experiments on flood forecasting have been collected or
produced when necessary using remote sensing data and integrated in a GIS application.

Task 10.3. Mapping and visualizing flood inundation, flood risk using combined satellite data and
hydraulic models (BNU, NIT)
The Chang Sha channel segment of Xiang Jiang River is selected as study area in Hunan Province of China.
The study area is located in the south of Dong Ting Lake. It is approximately located between 112 20 to
113 20 E and 27 50 to 28 30 N. The historical hydrologic data, including water level and
discharge of the observation station, have been collected for analysis of the development trends through the
time series analyzing method. In order to obtain precise DEM, we have bought 12 scenes of scale 1:10,000
topography maps in the study area, which have been digitized into ArcGIS vector format. Beside, some
remote sensing images, population data, and economic data of the study area have also been collected. All
these data are base for flood risk mapping study in the next step.

Significant results
A flood identification model has been made that includes two components:
flood risk mapping using soil moisture: ! (x, y, t)
flood risk mapping using soil moisture and local rainfall: ! (x, y, t), P (x, y, t)

AMSR-E soil moisture data provides time series of soil moisture. Satellite measurements of land surface
microwave emissivity such as AMSR-E describe the day-to-day variation of top soil moisture conditions.
Changes in the hydrological system are reflected in the soil moisture values. Land with suddenly rising soil
moisture values may be prone to flooding. Both absolute values and the time series of top soil moisture will
be used to provide a first identification of flood risk. Basically two aspects of soil moisture development in
time are important: (1) absolute soil moisture values in comparison to previous years; and (2) changes of soil
moisture in time (d!/dt).
Integration with rainfall data further refines the flood identification system. Rainfall data will be derived
from the Tropical Rainfall Measuring Mission (TRMM) satellite. TRMM rainfall provides information on
the location, quantity, intensity and timing of rainfall. TRMM rainfall is an extra source of information
measured independently of AMSR-E soil moisture for the flood model. The TRMM rainfall algorithm 3B42
(trmm.gsfc.nasa.gov/3b42.html) that provides 3-hourly rainfall data at 25 km resolution is most useful for
local rainfall and rainfall-runoff monitoring.
Rainfall observations fall into two components: local rainfall and rainfall in upstream catchments. Large
amounts of rainfall can cause flooding locally, but can also cause flooding downstream in the basin.

Figure 10.1 Bi-monthly soil moisture tolerance of two periods. When floods occurr the absolute moisture values exceed
these values.

Page 76 of 98


In China, work on preparing reference information to evaluate the new algorithm based on the polarization
difference in the 37GHz brightness temperature has concentrated on the Yangtze basin and the data collected
include for each flood the time of occurrence, intensity and location (Figure 10.2).

Figure 10.2 The locations of 10 major floods in Yangtze River basin since 1980; data on time of occurrence, intensity
and location are accessible through a GIS application.

In India, work on preparing reference information to evaluate the new algorithm based on the polarization
difference in the 37GHz brightness temperature has concentrated on the Ganga basin (Fig. 10.3). Collection
of the river flow data required significant efforts and are considered a critical information resource for the
next stages of the project.

Page 77 of 98


Fig.10.3 Mean flow, seasonal variation and lean flow in summer for selected basins in the Ganga Basin

The new XiAnJiang Model is being evaluated using data of the HuaiHe River in China (Figure 10.4).
Comparison of calculated with observed river flow iindicates a good agreement.

Figure 10.4. Forecast of floods of HuaiHe River in China with XiAnJiang Model; model vs, observed discharge show in
inset

Data collection for the case studies on flood propagation has been focusing on the Chang Sha channel
segment of Xiang Jiang River (Yangtze Basin), because this area is subject to frequent and severe floods and
the opportunity to understand better the driving factors of flooding in the catchment of the Dong Ting Lake
(Fig.10.5).

Page 78 of 98


Fig. 10.5 The Chang Sha channel segment of Xiang Jiang River in Hunan Province, China; the domain of the model to
describe flood propagation is shown (red).

Page 79 of 98


Summary of progress

This work-package started with some delay due to the difficulty to find a proper balance between work on
dissemination of project results through scientific events and in general international initiatives (e.g. GEO)
and true capacity buidling activities. In addition proper contacts with stakeholders organizations had to be
established, primarily in China. We had also underestimated the time and effort needed to make the CEOP –
AEGIS contribution to GEO visible. This all required a significant re-thinking of the work plan, timewise
rather than content – wise.

Task 11.1 Dissemination of project results
GEO: CEOP-AEGIS activities contribute to four societal benefits areas identified within GEOSS, ie. the
Reduction and Prevention of Disasters area, the Climate Change area, the Water Management area and the
Weather Forecasting area. In order to build up an observing system as a pilot of GEO system of systems for
water resources management, an important element is the dissemination of project information and results
and the involvement of stakeholders.
To successfully disseminate the knowledge gained in the project and make our contribution to GEOSS
visible, many activities are conducted in parallel:
! The organisation and contribution to international conferences and workshop
! The contribution to GEO activities
! The creation of communication media, ie. a website, paper and electronic brochures, multimedia
content
! The dissemination of knowledge through courses and training for representative stakeholders and for
young scientists
The work done and related outcome is described in detail in the Report De 11.1

Workshops and conferences.
Coordination meeting CEOP-AEGIS (CA) and CEOP-High Elevation (HE), June 29th – July 3rd 2009,
Hotel Ibis, Via Finocchiaro Aprile 2, 20124 MILANO, ITALY
This meeting was organized to establish cooperative links with research and capacity building program led
by CNR, Italy EvK2 CNR. This program has a permanent high elevation observatory in Nepal and is
carrying out interdisciplinary research and capacity building projects in the high elevation regions of Nepal
and Pakistan.

Task 11.2 Training sessions focus on human capacity building
The training program has been developed and it is included in De 11.1.
The objective of the advanced course is to train the participants from Asia of all the aspects related to in-situ
and earth observation, retrievals and modeling of land surface processes and land-atmosphere interactions
with emphasis on the Tibetan plateau. The course will include theory, instruments, validation, retrievals and
modeling and applications. Practical sessions will be oragnised with hands-on exercises with data collected
in different WPs. Lecturers are experts responsible for tasks in each WP.

Task 11.3 Tailored capacity building
1. Coordination meeting on Satellite based flood monitoring system of pilot areas of China and India, Indian
Institute of Technology Roorkee from September 12th to 14th 2009
This meeting was organized as a part of a UK – India Workshop on Water Resources Management under
Climate and Environment Change to inform a community focusing on hydrology and water management
about CEOP – AEGIS objectives and work.

Page 80 of 98


Significant results

Capacity Building web-page:
An element of the project web-site dedicated to dissemination and capacity building has been designed
(Fig.11.1).

Figure 11.1 Content of the Capacity Building pages on the CEOP-AEGIS website

This diagram summarizes the on-going dissemination, training and capacity building activities of the project.
The web site is also to be used to make widely available any output of the project relevant to capacity
building (see pages “Teaching and demonstration materials”).

Conferences
Overviews of CEOP – AEGIS objectives and progress were presented at the following International
Meetings:
- CEOP Implementation Meeting held in Geneva from the 15th to 18th Septembre 2008
- Reinforcing Europe’s contribution to, ISRSE-33 side event, May 5, 2009 PALAZZO DEI
CONGRESSI, STRESA, ITALY
- WCRP/GEWEX Melbourne, August 2009
These meetings had a considerable impact towards improving international awareness of CEOP – AEGIS
scope and establishing effective linkages with related international initiatives, relevant to CEOP – AEGIS

Page 81 of 98


from the point of view of either scientific issues, i.e. water cycle, land-atmosphere interactions and earth
observation or area of interest, i.e. SE Asia, or both.

The dissemination of information through the Internet being highly effective and in practice mandatory, the
CEOP-AEGIS Project Office bought the ceop-aegis.org domain name for an initial duration of 4 years. The
project website was constructed on an open source Web 2.0 Content Management System architecture call
MODx, and associated with a complete mailing list system. The domain name, the website and the mailing
lists rely on shared hardware plate-forms hosted by the University of Strasbourg. The website is designed to
host public information, news and material, as well as private content for registered users (ie. project

participants and stakeholders). The figure below shows the website map.

Figure: Diagram of the website structure. The green part is open to public access, while the orange and blue
ones are subject to registration.

The impact of the website can be approached by statistics on the activity on the www.ceop-aegis.org portal,
summarized in the following table.

Table: Statistics of www.ceop-aegis.org for 2009 and 2010.*for 2009 statistics were only available for the
last three months of the year (2nd version of the website).
Single visitor Visits Pages viewed Hits Bandwidth

2009* 333 812 24,596 60,498 498.67 Mo

2010 2,430 4,588 26,174 50,157 11.39Go

Page 82 of 98


Beside the website, a mailing list system is maintained by UDS to facilitate the communication within each
work-package, working-groups and lead scientific contacts. Moreover, efforts are made to encourage remote
communication through teleconference tools.

GEO Meetings
CEOP – AEGIS have been attending the following GEO meetings in the reporting period:
GEO CBC and STC Meeting Hannover
GEO European Project Workshops Bruxelles, Istanbul, Athens,

Page 83 of 98

CEOP-AEGIS Periodic Report #1

More Related Content

Viewers also liked

Similar to CEOP-AEGIS Periodic Report #1

More from jrgcolin

Recently uploaded

CEOP-AEGIS Periodic Report #1