Abstract image of a woman sitting on books and studying on a laptop

Grass gl30

Maria Antonia Brovelli, profile picture
Maria Antonia Brovelli

This document assesses land cover maps created with GRASS by comparing them to reference datasets using a confusion matrix approach. It describes the methodology for comparing maps and calculating accuracy statistics like overall accuracy. As a case study, it applies this process to compare the GlobeLand30 land cover maps for 2000 and 2010 to various maps of the Lombardy region of Italy. The results show overall accuracy between maps ranging from 81-92% without a buffer and 62-81% with a buffer to account for positional errors. Removing areas affected by positional errors increased overall accuracy.

Environment
1 of 51
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
Assessment of Land Coverage maps with GRASS Maria Antonia Brovelli, Monia Elisa Molinari GEOlab GEOlab, Politecnico di Milano – Como Campus, via Valleggio 11, 22100, Como, Italy 4th High Level Forum on Global Geospatial Information Management (UN-GGIM) Side Event: Globeland30 18-19 April 2016 Caucus 11, UNECA Conference Center Addis Ababa, Ethiopia
Index  Comparison methodology  Datasets  Data Processing  Case study: Lombardy Region (Italy) 2
A B C D A fAA fAB fAC fAD B fBA fBB fBC fBD C fCA fCB fCC fCD D fDA fDB fDC fDD CONFUSION MATRIX CLASSIFICATION The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one. GROUND TRUTH (REFERENCE) Comparison methodology 3
Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n CLASSIFICATION f+i = marginal sum of column i (i = A, B, C, D) fi+ = marginal sum of row i (i = A, B, C, D) n = total number of cells CONFUSION MATRIX GROUND TRUTH (REFERENCE) 4 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) Overall accuracy: evaluates the percentage of correctly classified pixels 𝑶𝑨 = 𝑓𝑖𝑖 𝑛 GROUND TRUTH (REFERENCE) 5 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) Allocation disagreement: the amount of difference due to the less then optimal match in the spatial allocation of the categories (Pontius and Millones, 2011) 𝑨𝑫 = (2 ∗ 𝑚𝑖𝑛 𝑓+𝑖 𝑛 − 𝑓𝑖𝑖 𝑛 , 𝑓𝑖+ 𝑛 − 𝑓𝑖𝑖 𝑛 ) 2 6 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS Quantity disagreement: the amount of difference due to the less than perfect match in the proportion of categories (Pontius and Millones, 2011) 𝑸𝑫 = 𝑓+𝑖 𝑛 − 𝑓𝑖+ 𝑛 2 (i = A, B, C, D) 7 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
8 Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS User’s accuracy: percentage of classified pixels which correctly match the ground truth 𝑼𝑨𝒊 = 𝑓𝑖𝑖 𝑓𝑖+ (i = A, B, C, D) The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
9 Comparison methodology Producer’s accuracy: percentage of pixels of ground truth correctly detected in the classified map 𝑷𝑨𝒊 = 𝑓𝑖𝑖 𝑓+𝑖 A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
10 Datasets: GlobeLand30 (GL30) UTM32 UTM33 • FORMAT: RASTER • YEARS: 2000, 2010 • REFERENCE SYSTEM: WGS84/UTM32 - WGS84/UTM33 Chen et al., 2014
11 Datasets: Italian Land Cover Data Friuli Venezia Giulia Land Cover 2000 - 1:25’000 Veneto Land Cover 2007/2009 - 1:10’000 Emilia Romagna Land Cover 2003 - 1:10’000 Land Cover 2008 - 1:10’000 Sardegna Land Cover 1997/2000 - 1:25’000 Land Cover 2003/2006 - 1:25’000 Liguria Land Cover 2000 - 1:10’000 Land Cover 2012 - 1:10’000 Lombardia Land Cover 1999/2000 - 1:10’000 Land Cover 2012 - 1:10’000 Autonomous Province of Bolzano Land Cover 2000 - 1:10’000 Autonomous Province of Trento Land Cover 2000 - 1:10’000 Abruzzo Land Cover 1997- 1:25’000
12 Data Processing REGION GL30 GL30
13 Data Processing RE-PROJECTION RE-PROJECTION REGION GL30 GL30 v.proj r.proj
14 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION MERGING v.proj r.proj r.patch
15 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast
16 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast RECLASSIFICATION RECLASSIFICATION r.reclass r.reclass CORINE LEGEND GLOBELAND30 LEGEND Artificial surfaces Artificial cover Agricultural areas Croplands Forests and semi natural areas Mixed forest, Broadleaf forest, Coniferous forest, Grass, Shrub, Bare land, Permanent ice or snow Wetlands Wetlands Water bodies Water
17 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast RECLASSIFICATION RECLASSIFICATION r.reclass r.reclass COMPARISON 1 r.kappa
18 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods
19 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods • Different classification schemes I METHOD II METHOD
20 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods • Different classification schemes • The co-location tolerance (Gallego, 2001) of GL30, which is equal to 70 m All cells belonging to a buffer of 70 m around GL30 classes border were eliminated and the confusion matrix and statistics were calculated on the other pixels. (Brovelli et al., 2015)
21 Lombardy Region case study DUSAF 1.1 DUSAF 2.0 DUSAF 2.1 DUSAF 3.0 DUSAF 4.0 YEAR 1999 - 2000 2005 - 2007 2007 2009 2012 SCALE 1:10’000 1:10’000 1:10’000 1:10’000 1:10’000 REF SYS WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N LEGEND CORINE CORINE CORINE CORINE CORINE SOURCE Aerial photos Aerial photos and regional databases Aerial photos and regional databases Aerial photos and regional databases Aerial photos and regional databases EXTENT whole region whole region whole region BS, MI, MB, SO, CR whole region DUSAF Land Cover Database DUSAF (“Use Categories of Agricultural and Forest Soils”) is a land cover database created in 2000-2001 within a project promoted and funded by Lombardy Region. The database provides a polygonal layer representing the land use and cover. Currently five releases are available.
22 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER
23 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER 30 m 90.53 7.40 2.07 30 m (prevalence) 90.64 7.33 2.03 5 m 90.83 7.10 2.07 • No significant changes in the results with different input dataset resolutions • The removal of the cells influenced by the co-location tolerance (buffer case) leads to an increase of the OA
24 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER 30 m 90.53 7.40 2.07 30 m (prevalence) 90.64 7.33 2.03 5 m 90.83 7.10 2.07 C1: artificial surfaces C2: Agricultural areas C3: Forests and semi natural areas C4: wetlands C5: water bodies PER-CLASS AGREEMENT MEASURES (30 m)
25 GL30 2000 – DUSAF1.1 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 77.18 17.97 4.85 30 m (prevalence) 77.30 17.84 4.85 5 m 77.13 18.02 4.85 BUFFER 30 m 82.03 12.82 5.15 30 m (prevalence) 82.15 12.74 5.10 5 m 82.43 12.36 5.21 9.2 6.3 2.85 Differences with respect the I method • The introduction of a greater level of classification detail entails an increase of both the allocation and quantity disagreement values.
26 GL30 2000 – DUSAF1.1 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 77.18 17.97 4.85 30 m (prevalence) 77.30 17.84 4.85 5 m 77.13 18.02 4.85 BUFFER 30 m 82.03 12.82 5.15 30 m (prevalence) 82.15 12.74 5.10 5 m 82.43 12.36 5.21 PER-CLASS AGREEMENT MEASURES (30 m) 9.2 6.3 2.85 Differences with respect the I method C31: forests C32: scrub and herbaceous vegetation C330: open space C335: glaciers /permanent snow
27 GL30 2010 – DUSAF4.0 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 79.88 16.12 4.00 30 m (prevalence) 80.01 16.09 3.89 5 m 79.83 16.17 4.00 BUFFER 30 m 85.06 10.74 4.19 30 m (prevalence) 85.20 10.68 4.20 5 m 85.57 10.19 4.24 PER-CLASS AGREEMENT MEASURES (30 m) 2.7 1.85 0.85 Differences with respect the year 2000 2000 value C31: forests C32: scrub and herbaceous vegetation C330: open space C335: glaciers /permanent snow
28 GL30 - Italian regions *OVERALL STATISTICS I METHOD - 2000 * No buffer 30 m resolution case *OVERALL STATISTICS I METHOD - 2010
29 GL30 - Italian regions *OVERALL STATISTICS II METHOD - 2010 *OVERALL STATISTICS II METHOD - 2000 * No buffer 30 m resolution case
30 Conclusions • The thematic accuracy assessment performed between GlobeLand30 and Italian regional land cover maps shows overall accuracy varying between: 81% - 92% (I method) 62% - 81% (II method) • The disagreement can be due to the fact that the images were taken on different dates, to the different thematic classification and resolution. • Removing the part of the disagreement due to the co-location tolerance the overall accuracy increases: 84% - 96 % (I method) 65% - 86% (II method) • In addition, in most of literature the reference data has been used as an accurate representation of the reality but they may contain errors.
31 References • Brovelli, M.A., Molinari, M.E., Hussein, E., Chen, J., Li, R. The First Comprehensive Accuracy Assessment of GlobeLand30 at a National Level: Methodology and Results. Remote Sens. 2015, 7(4), 4191-4212; doi:10.3390/rs70404191 • Chen, J.; Chen, J.; Liao, A.; Cao, X.; Chen, L.; Chen, X.; He, C.; Han, G.; Peng, S.; Lu, M.; et al. Global land cover mapping at 30 m resolution: A POK-based operational approach. ISPRS J. Photogram. Remote Sens. 2014 • Congalton, R.G.; Green, K. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices; Lewis Publishers: Boca Raton, FL, USA, 1999; p. 137 • Gallego, J. Comparing CORINE Land Cover with a more Detailed Database in Arezzo (Italy). Towards Agri-Environmental Indicators; Topic report 6/2001 European Environment Agency 2001; European Environment Agency: Copenhagen, Danmark, 2001; pp. 118–125 • GRASS Development Team, 2015. Geographic Resources Analysis Support System (GRASS) Software, Version 6.4.5. Open Source Geospatial Foundation. http://grass.osgeo.org • Pontius, R.G.; Millones, M. Death to Kappa: Birth of quantity disagreement and allocation disagreement for accuracy assessment. Int. J. Remote Sens. 2011, 32, 4407– 4429
32 What is GRASS GIS GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Geographic Resources Analysis Support System • 1982: first developments carried out at USA/CERL (Construction Engineering Research Laboratory) • 1985: version 1.0 was released • 1998: development transferred to an international GRASS Development Team led by Markus Neteler • 1999: GRASS 5.0 release under General Public License (GPL) • 2016: stable long term release GRASS 7.0.3
33 Download GRASS GIS GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ https://grass.osgeo.org/download/
34 GEOlab, Politecnico di Milano – Como Campus Starting with GRASS GIS 1. GIS Database directory: directory containing all the GRASS GIS data 2. Location: it is a sort of data library for the region of interest. It contains datasets with the same coordinate system 3. Mapset: it is a location subdirectory where GIS maps are organized thematically or geographically DATABASE LOCATION 1 LOCATION 2 LOCATION N PERMANENT MAPSET 1 MAPSET N MULTI-USER
35 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name of the location
36 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
37 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
38 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
39 GEOlab, Politecnico di Milano – Como Campus Mapset Once the location is created, the system requires information about:  region extents  mapset name
40 GEOlab, Politecnico di Milano – Como Campus GRASS GIS interface Layer Manager  Toolbar to manage displayed map layers  Menu bar with all GRASS GIS functions  Python shell Map Display  2D and 3D view of the maps  Tools for map navigation  Tools for map analysis  Map elements  Tools for export and printing Bash Shell  It runs GRASS GIS modules without GUI
41 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ GRASS GIS modules PREFIX CLASS FUNCTION d.* display Visualization db.* database Database management g.* general General file operations i.* image Image processing ps.* postscript Map creation in Postscript r.* raster Raster analysis r3.* voxel Voxel analysis v.* vector Vector analysis t.* timeseries Temporal data processing m.* miscellaneous Miscellaneous functions GRASS is a very powerful GIS suite with over than 400 standard modules in the core version. https://grass.osgeo.org/grass70/manuals/index.html
42 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Data import: v.in.ogr v.in.ogr module converts OGR vector layers to GRASS vector map. https://grass.osgeo.org/grass70/manuals/v.in.ogr
43 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Data import: r.in.gdal r.in.gdal module imports GDAL supported raster file into a binary raster map layer. https://grass.osgeo.org/grass70/manuals/r.in.gdal
44 GEOlab, Politecnico di Milano – Como Campus GRASS region g.region module manages the boundary definitions for the geographic region. The region is a geographic area with defined boundaries (N,W,S,E) and specific E-W and N-S resolutions of its smallest units (cells). Most of the raster and displays modules are affected by the current region settings. https://grass.osgeo.org/grass70/manuals/g.region Define the current region specifying coordinates and resolutions
45 GEOlab, Politecnico di Milano – Como Campus GRASS region g.region module manages the boundary definitions for the geographic region. The region is a geographic area with defined boundaries (N,W,S,E) and specific E-W and N-S resolutions of its smallest units (cells). Most of the raster and displays modules are affected by the current region settings. Define the current region settings to match the extension of a raster/vector map https://grass.osgeo.org/grass70/manuals/g.region
46 GEOlab, Politecnico di Milano – Como Campus Re-projection: v.proj / r.proj v.proj / r.proj modules re-projects a vector / raster map from one location to the current location. 1. Create a location in the layer reference system 2. Import the layer into the location 3. Create the location in the reference system you want to re-project the layer into 4. Run v.proj (for vector map) or r.proj (for raster map) modules https://grass.osgeo.org/grass70/manuals/v.proj.html https://grass.osgeo.org/grass70/manuals/r.proj.html
47 GEOlab, Politecnico di Milano – Como Campus Merging: r.patch r.patch module creates a composite raster map layer by using known category values from one (or more) map layer(s) to fill in areas of "no data" in another map layer. https://grass.osgeo.org/grass70/manuals/r.patch.html
48 GEOlab, Politecnico di Milano – Como Campus Rasterization: v.to.rast v.to.rast converts a vector map into a raster map. The resolution of the output raster map is the same of the current region. https://grass.osgeo.org/grass70/manuals/v.to.rast.html
49 GEOlab, Politecnico di Milano – Como Campus Reclassification: r.reclass r.reclass reclassifies a raster map based on category values. https://grass.osgeo.org/grass70/manuals/r.reclass.html
50 GEOlab, Politecnico di Milano – Como Campus Matrix confusion calculation: r.kappa r.kappa calculates error matrix for accuracy assessment of classification result. https://grass.osgeo.org/grass70/manuals/r.kappa.html
51 GEOlab, Politecnico di Milano – Como Campus Matrix confusion calculation: r.kappa r.kappa calculates error matrix for accuracy assessment of classification result. https://grass.osgeo.org/grass70/manuals/r.kappa.html Stats.txt Confusion matrix Commission and omission errors Overall accuracy

More Related Content

PDF
Exploring DEM error with geographically weighted regression
GeoCommunity
 
PPT
20110723IGARSS_ZHAO-yang.ppt
grssieee
 
PDF
Velocity model building in Petrel
Amir Abbas Babasafari
 
DOCX
Applied Geospatial Sciences Assignment
Joshua Brunsdon
 
PDF
ESPL1351
Joshua Brunsdon
 
PPT
Interpolation of meteodata using the method of regression-kriging
Alexander Mkrtchian
 
PPTX
Petrel course Module_1: Import data and management, make simple surfaces
Marc Diviu Franco
 
PDF
3D Visualizations of Land Cover Maps
IOSRjournaljce
 
Exploring DEM error with geographically weighted regression
GeoCommunity
 
20110723IGARSS_ZHAO-yang.ppt
grssieee
 
Velocity model building in Petrel
Amir Abbas Babasafari
 
Applied Geospatial Sciences Assignment
Joshua Brunsdon
 
ESPL1351
Joshua Brunsdon
 
Interpolation of meteodata using the method of regression-kriging
Alexander Mkrtchian
 
Petrel course Module_1: Import data and management, make simple surfaces
Marc Diviu Franco
 
3D Visualizations of Land Cover Maps
IOSRjournaljce
 

What's hot (20)

PPT
Geoland2 Workshop - 01 - sousa
Planetek Italia Srl
 
PDF
B1080817
IJERD Editor
 
PDF
Belmont Res.Kibby Basin,NV. Property Gravity Survey Basin Model
Follow me on Twitter @Stockshaman
 
PDF
Using Artificial Neural Networks for Digital Soil Mapping – a comparison of M...
Ricardo Brasil
 
PPS
Structure-metric method FOR PREDICTIVE ESTIMATION of NATURAL RESOURCES
KaterinaKaritskaya
 
PDF
A statistical assessment of GDEM using LiDAR data
Tomislav Hengl
 
PDF
Spatial analysis for the assessment of the environmental changes in the lands...
Universität Salzburg
 
PDF
Spatial interpolation comparison
Tomislav Hengl
 
PDF
APPLICATION OF KRIGING IN GROUND WATER STUDIES
Abhiram Kanigolla
 
PDF
Using Kriging Combined with Categorical Information of Soil Maps for Interpol...
FAO
 
PPTX
Static model development
Kunal Rathod
 
PDF
MODELAÇÃO DO SOLO-PAISAGEM – A IMPORTÂNCIA DA LOCALIZAÇÃO
Ricardo Brasil
 
PDF
Petrel introduction course guide
Marc Diviu Franco
 
PDF
Assessing 50 years of tropical Peruvian glacier volume change from multitempo...
InfoAndina CONDESAN
 
PPT
Urban Land Cover Change Detection Analysis and Modelling Spatio-Temporal Grow...
Bayes Ahmed
 
PDF
User guide of reservoir geological modeling v2.2.0
Bo Sun
 
PDF
2. virat arora, s. s. rao, e. amminedu and p. jagadeeswara rao
Journal of Global Resources
 
PPTX
Thesis defense
Eridon Brahimllari
 
PDF
7th euregeo volume_1 158_159
Ricardo Brasil
 
PDF
Settlement prediction research on the gravel pile in soft soil subgrade
IJERA Editor
 
Geoland2 Workshop - 01 - sousa
Planetek Italia Srl
 
B1080817
IJERD Editor
 
Belmont Res.Kibby Basin,NV. Property Gravity Survey Basin Model
Follow me on Twitter @Stockshaman
 
Using Artificial Neural Networks for Digital Soil Mapping – a comparison of M...
Ricardo Brasil
 
Structure-metric method FOR PREDICTIVE ESTIMATION of NATURAL RESOURCES
KaterinaKaritskaya
 
A statistical assessment of GDEM using LiDAR data
Tomislav Hengl
 
Spatial analysis for the assessment of the environmental changes in the lands...
Universität Salzburg
 
Spatial interpolation comparison
Tomislav Hengl
 
APPLICATION OF KRIGING IN GROUND WATER STUDIES
Abhiram Kanigolla
 
Using Kriging Combined with Categorical Information of Soil Maps for Interpol...
FAO
 
Static model development
Kunal Rathod
 
MODELAÇÃO DO SOLO-PAISAGEM – A IMPORTÂNCIA DA LOCALIZAÇÃO
Ricardo Brasil
 
Petrel introduction course guide
Marc Diviu Franco
 
Assessing 50 years of tropical Peruvian glacier volume change from multitempo...
InfoAndina CONDESAN
 
Urban Land Cover Change Detection Analysis and Modelling Spatio-Temporal Grow...
Bayes Ahmed
 
User guide of reservoir geological modeling v2.2.0
Bo Sun
 
2. virat arora, s. s. rao, e. amminedu and p. jagadeeswara rao
Journal of Global Resources
 
Thesis defense
Eridon Brahimllari
 
7th euregeo volume_1 158_159
Ricardo Brasil
 
Settlement prediction research on the gravel pile in soft soil subgrade
IJERA Editor
 
Ad

Similar to Grass gl30 (20)

PDF
Land cover game lps16.ppt
Maria Antonia Brovelli
 
PPTX
Improving input data for urban canopy and land surface models: a sensitivity ...
National Institute of Food and Agriculture
 
PDF
DRONES IN HYDROLOGY
Salvatore Manfreda
 
PDF
Regional-Scale Assessment of N2O Emissions within the US Corn Belt: The Impac...
National Institute of Food and Agriculture
 
PDF
FOSS4G in Europe; Italy and the Politecnico de Milano
Carolina Arias Muñoz
 
PDF
Anàlisi de les afectacions de la processionària amb UAV i LANDSAT
ICGCat
 
PPTX
Kim_WE3_T05_2.pptx
grssieee
 
PPTX
GSOC17 Introduction, Product specifications, Existing SOC maps and methodologies
FAO
 
PDF
Tarımsal Toprak Haritalama'da Jeofizik Mühendisliği
Ali Osman Öncel
 
PDF
New swat tile drain equations
Soil and Water Conservation Society
 
PPT
Cn tu12 6_isric_glada_mapping_of_desire_study_sites_bai
Erik van den Elsen
 
PPT
Bouvet_IGARSS2011.ppt
grssieee
 
PDF
Assessment of Urban Green Space Structures and Its Effect on Land Surface Tem...
Manat Srivanit
 
PPTX
Automated change detection in grass gis
COGS Presentations
 
PPTX
Soil Organic Carbon Map of Mexico
ExternalEvents
 
PDF
Costa, Ciniro Jr - Climate Food and Farming CLIFF Network annual workshop Nov...
CCAFS | CGIAR Research Program on Climate Change, Agriculture and Food Security
 
PDF
Enhancing Landscape Recolonization Assessment Through FME and Big Data Integr...
Safe Software
 
PPTX
Generating Training Data from Noisy Measrements
Louisa Diggs
 
PPT
A strategy for estimating the components of the carbon and water budgets for ...
Integrated Carbon Observation System (ICOS)
 
PDF
UAS based soil moisture monitoring
Salvatore Manfreda
 
Land cover game lps16.ppt
Maria Antonia Brovelli
 
Improving input data for urban canopy and land surface models: a sensitivity ...
National Institute of Food and Agriculture
 
DRONES IN HYDROLOGY
Salvatore Manfreda
 
Regional-Scale Assessment of N2O Emissions within the US Corn Belt: The Impac...
National Institute of Food and Agriculture
 
FOSS4G in Europe; Italy and the Politecnico de Milano
Carolina Arias Muñoz
 
Anàlisi de les afectacions de la processionària amb UAV i LANDSAT
ICGCat
 
Kim_WE3_T05_2.pptx
grssieee
 
GSOC17 Introduction, Product specifications, Existing SOC maps and methodologies
FAO
 
Tarımsal Toprak Haritalama'da Jeofizik Mühendisliği
Ali Osman Öncel
 
New swat tile drain equations
Soil and Water Conservation Society
 
Cn tu12 6_isric_glada_mapping_of_desire_study_sites_bai
Erik van den Elsen
 
Bouvet_IGARSS2011.ppt
grssieee
 
Assessment of Urban Green Space Structures and Its Effect on Land Surface Tem...
Manat Srivanit
 
Automated change detection in grass gis
COGS Presentations
 
Soil Organic Carbon Map of Mexico
ExternalEvents
 
Costa, Ciniro Jr - Climate Food and Farming CLIFF Network annual workshop Nov...
CCAFS | CGIAR Research Program on Climate Change, Agriculture and Food Security
 
Enhancing Landscape Recolonization Assessment Through FME and Big Data Integr...
Safe Software
 
Generating Training Data from Noisy Measrements
Louisa Diggs
 
A strategy for estimating the components of the carbon and water budgets for ...
Integrated Carbon Observation System (ICOS)
 
UAS based soil moisture monitoring
Salvatore Manfreda
 
Ad

More from Maria Antonia Brovelli (20)

PPTX
Academic Network Strategy on implementing UN IGIF
Maria Antonia Brovelli
 
PDF
Side Event 31 July - Academic Network.pdf
Maria Antonia Brovelli
 
PPTX
ANALYSIS OF THE REMOTELY SENSED WATER QUALITY PARAMETERS OF THE INSUBRIC LAKE...
Maria Antonia Brovelli
 
PDF
ISDE 2023 Brovelli.pdf
Maria Antonia Brovelli
 
PDF
GIS e Climate Change nelle scuole superiori in Europa: il progetto GIS4Schools
Maria Antonia Brovelli
 
PDF
GIS4Schools
Maria Antonia Brovelli
 
PDF
Mobile tools for community scientists
Maria Antonia Brovelli
 
PDF
Open technologies for United Nations
Maria Antonia Brovelli
 
PDF
Global Land Cover and Intelligent Analysis of Remote Sensed Images
Maria Antonia Brovelli
 
PDF
EO in Society: Open Science and Innovation
Maria Antonia Brovelli
 
PDF
Smart Citizens
Maria Antonia Brovelli
 
PDF
Smart Citizens
Maria Antonia Brovelli
 
PDF
UN Open GIS Capacity Building
Maria Antonia Brovelli
 
PDF
From OpenStreetMap data to Land Use/Land Cover maps: experiments and first re...
Maria Antonia Brovelli
 
PDF
Lesson3 esa summer_school_brovelli
Maria Antonia Brovelli
 
PDF
Lesson2 esa summer_school_brovelli
Maria Antonia Brovelli
 
PDF
Lesson1 esa summer_school_brovelli
Maria Antonia Brovelli
 
PDF
Lps16 nasa
Maria Antonia Brovelli
 
PDF
Minimapathons
Maria Antonia Brovelli
 
PDF
GeoForAll: a successful OSGeo Initiative
Maria Antonia Brovelli
 
Academic Network Strategy on implementing UN IGIF
Maria Antonia Brovelli
 
Side Event 31 July - Academic Network.pdf
Maria Antonia Brovelli
 
ANALYSIS OF THE REMOTELY SENSED WATER QUALITY PARAMETERS OF THE INSUBRIC LAKE...
Maria Antonia Brovelli
 
ISDE 2023 Brovelli.pdf
Maria Antonia Brovelli
 
GIS e Climate Change nelle scuole superiori in Europa: il progetto GIS4Schools
Maria Antonia Brovelli
 
GIS4Schools
Maria Antonia Brovelli
 
Mobile tools for community scientists
Maria Antonia Brovelli
 
Open technologies for United Nations
Maria Antonia Brovelli
 
Global Land Cover and Intelligent Analysis of Remote Sensed Images
Maria Antonia Brovelli
 
EO in Society: Open Science and Innovation
Maria Antonia Brovelli
 
Smart Citizens
Maria Antonia Brovelli
 
Smart Citizens
Maria Antonia Brovelli
 
UN Open GIS Capacity Building
Maria Antonia Brovelli
 
From OpenStreetMap data to Land Use/Land Cover maps: experiments and first re...
Maria Antonia Brovelli
 
Lesson3 esa summer_school_brovelli
Maria Antonia Brovelli
 
Lesson2 esa summer_school_brovelli
Maria Antonia Brovelli
 
Lesson1 esa summer_school_brovelli
Maria Antonia Brovelli
 
Lps16 nasa
Maria Antonia Brovelli
 
Minimapathons
Maria Antonia Brovelli
 
GeoForAll: a successful OSGeo Initiative
Maria Antonia Brovelli
 

Recently uploaded (20)

DOCX
The Ripple Effect: Understanding Extreme Weather Patterns and Geomagnetic Dyn...
kutatomoshi
 
PPTX
IMPACTS OF CHEMICAL FERTILIZERS -CHEMPROJ (11).pptx
Saksham Manandhar
 
PPTX
EME Aerospace.pptx basics of mechanical engineering
AMAN55070
 
PPTX
Drought and it's classification in a simplified way
satyaprakashstar17
 
PPTX
Climate_Change_Renewable_and_Energy.pptx
sufidelio
 
PDF
PAKAM TECHNOLOGY LIMTED PITCH DECK pptx.pdf
leyeode1980
 
PPTX
Pollution, it's Types and Impacts on Global context.pptx
brilliantanshul
 
PDF
The European Green Deal (EU Green Deal)
Naeemshahzad51
 
DOCX
Aluminum Dome Roofs for Agricultural Digesters A Durable Cover for Structural...
AllenLin596164
 
PPTX
14.1 Opinion Essay (Writing). to teach opinion
Doukanimek5
 
PPTX
RadiationSafetyPt120252026nucchemis.pptx
ChrisPatron4
 
PPTX
Psychological Support for Elderly During Health Crises.pptx
mdgibuswa
 
PPTX
Advances in Integrated Nutrient and Insect-pest Management in Vegetable Crops...
manasouat15
 
PPTX
APR 05.05.25.pptx gffdtkdtxfxtdytdtdcfcfxr
ppdsbedahfeb25
 
PPTX
Unit 1 - Environmental management, politics and.pptx
mishrasabi333
 
DOCX
Biogas Tank for Bio CNG Plants The main container for biogas digestion..docx
AllenLin596164
 
PDF
IWRM - City University Presentation 28 may 2018-v3.pdf
halwani1
 
PPTX
-Case-Study 2 What Happened in the Cavite Mutiny (1).pptx
batocaelareballecass
 
PDF
Biomass cookstoves: A review of technical aspects
AhmadSubiyanto
 
PDF
BD4E4- DISASTER MANAGEMENT BY A.R.SIVANESH.pdf
A R SIVANESH M.E., (Ph.D)
 
The Ripple Effect: Understanding Extreme Weather Patterns and Geomagnetic Dyn...
kutatomoshi
 
IMPACTS OF CHEMICAL FERTILIZERS -CHEMPROJ (11).pptx
Saksham Manandhar
 
EME Aerospace.pptx basics of mechanical engineering
AMAN55070
 
Drought and it's classification in a simplified way
satyaprakashstar17
 
Climate_Change_Renewable_and_Energy.pptx
sufidelio
 
PAKAM TECHNOLOGY LIMTED PITCH DECK pptx.pdf
leyeode1980
 
Pollution, it's Types and Impacts on Global context.pptx
brilliantanshul
 
The European Green Deal (EU Green Deal)
Naeemshahzad51
 
Aluminum Dome Roofs for Agricultural Digesters A Durable Cover for Structural...
AllenLin596164
 
14.1 Opinion Essay (Writing). to teach opinion
Doukanimek5
 
RadiationSafetyPt120252026nucchemis.pptx
ChrisPatron4
 
Psychological Support for Elderly During Health Crises.pptx
mdgibuswa
 
Advances in Integrated Nutrient and Insect-pest Management in Vegetable Crops...
manasouat15
 
APR 05.05.25.pptx gffdtkdtxfxtdytdtdcfcfxr
ppdsbedahfeb25
 
Unit 1 - Environmental management, politics and.pptx
mishrasabi333
 
Biogas Tank for Bio CNG Plants The main container for biogas digestion..docx
AllenLin596164
 
IWRM - City University Presentation 28 may 2018-v3.pdf
halwani1
 
-Case-Study 2 What Happened in the Cavite Mutiny (1).pptx
batocaelareballecass
 
Biomass cookstoves: A review of technical aspects
AhmadSubiyanto
 
BD4E4- DISASTER MANAGEMENT BY A.R.SIVANESH.pdf
A R SIVANESH M.E., (Ph.D)
 

Grass gl30

  • 1. Assessment of Land Coverage maps with GRASS Maria Antonia Brovelli, Monia Elisa Molinari GEOlab GEOlab, Politecnico di Milano – Como Campus, via Valleggio 11, 22100, Como, Italy 4th High Level Forum on Global Geospatial Information Management (UN-GGIM) Side Event: Globeland30 18-19 April 2016 Caucus 11, UNECA Conference Center Addis Ababa, Ethiopia
  • 2. Index  Comparison methodology  Datasets  Data Processing  Case study: Lombardy Region (Italy) 2
  • 3. A B C D A fAA fAB fAC fAD B fBA fBB fBC fBD C fCA fCB fCC fCD D fDA fDB fDC fDD CONFUSION MATRIX CLASSIFICATION The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one. GROUND TRUTH (REFERENCE) Comparison methodology 3
  • 4. Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n CLASSIFICATION f+i = marginal sum of column i (i = A, B, C, D) fi+ = marginal sum of row i (i = A, B, C, D) n = total number of cells CONFUSION MATRIX GROUND TRUTH (REFERENCE) 4 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 5. Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) Overall accuracy: evaluates the percentage of correctly classified pixels 𝑶𝑨 = 𝑓𝑖𝑖 𝑛 GROUND TRUTH (REFERENCE) 5 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 6. Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) Allocation disagreement: the amount of difference due to the less then optimal match in the spatial allocation of the categories (Pontius and Millones, 2011) 𝑨𝑫 = (2 ∗ 𝑚𝑖𝑛 𝑓+𝑖 𝑛 − 𝑓𝑖𝑖 𝑛 , 𝑓𝑖+ 𝑛 − 𝑓𝑖𝑖 𝑛 ) 2 6 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 7. Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS Quantity disagreement: the amount of difference due to the less than perfect match in the proportion of categories (Pontius and Millones, 2011) 𝑸𝑫 = 𝑓+𝑖 𝑛 − 𝑓𝑖+ 𝑛 2 (i = A, B, C, D) 7 The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 8. 8 Comparison methodology A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS User’s accuracy: percentage of classified pixels which correctly match the ground truth 𝑼𝑨𝒊 = 𝑓𝑖𝑖 𝑓𝑖+ (i = A, B, C, D) The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 9. 9 Comparison methodology Producer’s accuracy: percentage of pixels of ground truth correctly detected in the classified map 𝑷𝑨𝒊 = 𝑓𝑖𝑖 𝑓+𝑖 A B C D A fAA fAB fAC fAD fA+ B fBA fBB fBC fBD fB+ C fCA fCB fCC fCD fC+ D fDA fDB fDC fDD fD+ f+A f+B f+C f+D n GROUND TRUTH (REFERENCE) CLASSIFICATION CONFUSION MATRIX DERIVED STATISTICS (i = A, B, C, D) The confusion matrix (Congalton and Green, 1999) is derived through a pixel by pixel comparison between a classified map and a reference one.
  • 10. 10 Datasets: GlobeLand30 (GL30) UTM32 UTM33 • FORMAT: RASTER • YEARS: 2000, 2010 • REFERENCE SYSTEM: WGS84/UTM32 - WGS84/UTM33 Chen et al., 2014
  • 11. 11 Datasets: Italian Land Cover Data Friuli Venezia Giulia Land Cover 2000 - 1:25’000 Veneto Land Cover 2007/2009 - 1:10’000 Emilia Romagna Land Cover 2003 - 1:10’000 Land Cover 2008 - 1:10’000 Sardegna Land Cover 1997/2000 - 1:25’000 Land Cover 2003/2006 - 1:25’000 Liguria Land Cover 2000 - 1:10’000 Land Cover 2012 - 1:10’000 Lombardia Land Cover 1999/2000 - 1:10’000 Land Cover 2012 - 1:10’000 Autonomous Province of Bolzano Land Cover 2000 - 1:10’000 Autonomous Province of Trento Land Cover 2000 - 1:10’000 Abruzzo Land Cover 1997- 1:25’000
  • 15. 15 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast
  • 16. 16 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast RECLASSIFICATION RECLASSIFICATION r.reclass r.reclass CORINE LEGEND GLOBELAND30 LEGEND Artificial surfaces Artificial cover Agricultural areas Croplands Forests and semi natural areas Mixed forest, Broadleaf forest, Coniferous forest, Grass, Shrub, Bare land, Permanent ice or snow Wetlands Wetlands Water bodies Water
  • 17. 17 Data Processing REGION GL30 GL30 RE-PROJECTION RE-PROJECTION v.proj r.proj MERGING r.patch RASTERIZATION v.to.rast RECLASSIFICATION RECLASSIFICATION r.reclass r.reclass COMPARISON 1 r.kappa
  • 18. 18 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods
  • 19. 19 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods • Different classification schemes I METHOD II METHOD
  • 20. 20 Lombardy Region case study The procedure has been applied taking into account the influence on the comparison results of: • Different rasterization resolutions and methods • Different classification schemes • The co-location tolerance (Gallego, 2001) of GL30, which is equal to 70 m All cells belonging to a buffer of 70 m around GL30 classes border were eliminated and the confusion matrix and statistics were calculated on the other pixels. (Brovelli et al., 2015)
  • 21. 21 Lombardy Region case study DUSAF 1.1 DUSAF 2.0 DUSAF 2.1 DUSAF 3.0 DUSAF 4.0 YEAR 1999 - 2000 2005 - 2007 2007 2009 2012 SCALE 1:10’000 1:10’000 1:10’000 1:10’000 1:10’000 REF SYS WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N WGS84/UTM32N LEGEND CORINE CORINE CORINE CORINE CORINE SOURCE Aerial photos Aerial photos and regional databases Aerial photos and regional databases Aerial photos and regional databases Aerial photos and regional databases EXTENT whole region whole region whole region BS, MI, MB, SO, CR whole region DUSAF Land Cover Database DUSAF (“Use Categories of Agricultural and Forest Soils”) is a land cover database created in 2000-2001 within a project promoted and funded by Lombardy Region. The database provides a polygonal layer representing the land use and cover. Currently five releases are available.
  • 22. 22 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER
  • 23. 23 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER 30 m 90.53 7.40 2.07 30 m (prevalence) 90.64 7.33 2.03 5 m 90.83 7.10 2.07 • No significant changes in the results with different input dataset resolutions • The removal of the cells influenced by the co-location tolerance (buffer case) leads to an increase of the OA
  • 24. 24 GL30 2000 – DUSAF1.1 OVERALL STATISTICS I METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 86.39 11.61 2.00 30 m (prevalence) 86.50 11.52 1.98 5 m 86.34 11.66 2.00 BUFFER 30 m 90.53 7.40 2.07 30 m (prevalence) 90.64 7.33 2.03 5 m 90.83 7.10 2.07 C1: artificial surfaces C2: Agricultural areas C3: Forests and semi natural areas C4: wetlands C5: water bodies PER-CLASS AGREEMENT MEASURES (30 m)
  • 25. 25 GL30 2000 – DUSAF1.1 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 77.18 17.97 4.85 30 m (prevalence) 77.30 17.84 4.85 5 m 77.13 18.02 4.85 BUFFER 30 m 82.03 12.82 5.15 30 m (prevalence) 82.15 12.74 5.10 5 m 82.43 12.36 5.21 9.2 6.3 2.85 Differences with respect the I method • The introduction of a greater level of classification detail entails an increase of both the allocation and quantity disagreement values.
  • 26. 26 GL30 2000 – DUSAF1.1 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 77.18 17.97 4.85 30 m (prevalence) 77.30 17.84 4.85 5 m 77.13 18.02 4.85 BUFFER 30 m 82.03 12.82 5.15 30 m (prevalence) 82.15 12.74 5.10 5 m 82.43 12.36 5.21 PER-CLASS AGREEMENT MEASURES (30 m) 9.2 6.3 2.85 Differences with respect the I method C31: forests C32: scrub and herbaceous vegetation C330: open space C335: glaciers /permanent snow
  • 27. 27 GL30 2010 – DUSAF4.0 OVERALL STATISTICS II METHOD RESOLUTION OA [%] AD [%] QD [%] NO BUFFER 30 m 79.88 16.12 4.00 30 m (prevalence) 80.01 16.09 3.89 5 m 79.83 16.17 4.00 BUFFER 30 m 85.06 10.74 4.19 30 m (prevalence) 85.20 10.68 4.20 5 m 85.57 10.19 4.24 PER-CLASS AGREEMENT MEASURES (30 m) 2.7 1.85 0.85 Differences with respect the year 2000 2000 value C31: forests C32: scrub and herbaceous vegetation C330: open space C335: glaciers /permanent snow
  • 28. 28 GL30 - Italian regions *OVERALL STATISTICS I METHOD - 2000 * No buffer 30 m resolution case *OVERALL STATISTICS I METHOD - 2010
  • 29. 29 GL30 - Italian regions *OVERALL STATISTICS II METHOD - 2010 *OVERALL STATISTICS II METHOD - 2000 * No buffer 30 m resolution case
  • 30. 30 Conclusions • The thematic accuracy assessment performed between GlobeLand30 and Italian regional land cover maps shows overall accuracy varying between: 81% - 92% (I method) 62% - 81% (II method) • The disagreement can be due to the fact that the images were taken on different dates, to the different thematic classification and resolution. • Removing the part of the disagreement due to the co-location tolerance the overall accuracy increases: 84% - 96 % (I method) 65% - 86% (II method) • In addition, in most of literature the reference data has been used as an accurate representation of the reality but they may contain errors.
  • 31. 31 References • Brovelli, M.A., Molinari, M.E., Hussein, E., Chen, J., Li, R. The First Comprehensive Accuracy Assessment of GlobeLand30 at a National Level: Methodology and Results. Remote Sens. 2015, 7(4), 4191-4212; doi:10.3390/rs70404191 • Chen, J.; Chen, J.; Liao, A.; Cao, X.; Chen, L.; Chen, X.; He, C.; Han, G.; Peng, S.; Lu, M.; et al. Global land cover mapping at 30 m resolution: A POK-based operational approach. ISPRS J. Photogram. Remote Sens. 2014 • Congalton, R.G.; Green, K. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices; Lewis Publishers: Boca Raton, FL, USA, 1999; p. 137 • Gallego, J. Comparing CORINE Land Cover with a more Detailed Database in Arezzo (Italy). Towards Agri-Environmental Indicators; Topic report 6/2001 European Environment Agency 2001; European Environment Agency: Copenhagen, Danmark, 2001; pp. 118–125 • GRASS Development Team, 2015. Geographic Resources Analysis Support System (GRASS) Software, Version 6.4.5. Open Source Geospatial Foundation. http://grass.osgeo.org • Pontius, R.G.; Millones, M. Death to Kappa: Birth of quantity disagreement and allocation disagreement for accuracy assessment. Int. J. Remote Sens. 2011, 32, 4407– 4429
  • 32. 32 What is GRASS GIS GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Geographic Resources Analysis Support System • 1982: first developments carried out at USA/CERL (Construction Engineering Research Laboratory) • 1985: version 1.0 was released • 1998: development transferred to an international GRASS Development Team led by Markus Neteler • 1999: GRASS 5.0 release under General Public License (GPL) • 2016: stable long term release GRASS 7.0.3
  • 33. 33 Download GRASS GIS GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ https://grass.osgeo.org/download/
  • 34. 34 GEOlab, Politecnico di Milano – Como Campus Starting with GRASS GIS 1. GIS Database directory: directory containing all the GRASS GIS data 2. Location: it is a sort of data library for the region of interest. It contains datasets with the same coordinate system 3. Mapset: it is a location subdirectory where GIS maps are organized thematically or geographically DATABASE LOCATION 1 LOCATION 2 LOCATION N PERMANENT MAPSET 1 MAPSET N MULTI-USER
  • 35. 35 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name of the location
  • 36. 36 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
  • 37. 37 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
  • 38. 38 GEOlab, Politecnico di Milano – Como Campus Location To create a Location click on New button in the GRASS GIS startup window. Then, in the GRASS GIS Location window define:  The path to the GIS database directory  The name for the project location  The reference system
  • 39. 39 GEOlab, Politecnico di Milano – Como Campus Mapset Once the location is created, the system requires information about:  region extents  mapset name
  • 40. 40 GEOlab, Politecnico di Milano – Como Campus GRASS GIS interface Layer Manager  Toolbar to manage displayed map layers  Menu bar with all GRASS GIS functions  Python shell Map Display  2D and 3D view of the maps  Tools for map navigation  Tools for map analysis  Map elements  Tools for export and printing Bash Shell  It runs GRASS GIS modules without GUI
  • 41. 41 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ GRASS GIS modules PREFIX CLASS FUNCTION d.* display Visualization db.* database Database management g.* general General file operations i.* image Image processing ps.* postscript Map creation in Postscript r.* raster Raster analysis r3.* voxel Voxel analysis v.* vector Vector analysis t.* timeseries Temporal data processing m.* miscellaneous Miscellaneous functions GRASS is a very powerful GIS suite with over than 400 standard modules in the core version. https://grass.osgeo.org/grass70/manuals/index.html
  • 42. 42 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Data import: v.in.ogr v.in.ogr module converts OGR vector layers to GRASS vector map. https://grass.osgeo.org/grass70/manuals/v.in.ogr
  • 43. 43 GEOlab, Politecnico di Milano – Como Campus http://www.openstreetmap.org/ Data import: r.in.gdal r.in.gdal module imports GDAL supported raster file into a binary raster map layer. https://grass.osgeo.org/grass70/manuals/r.in.gdal
  • 44. 44 GEOlab, Politecnico di Milano – Como Campus GRASS region g.region module manages the boundary definitions for the geographic region. The region is a geographic area with defined boundaries (N,W,S,E) and specific E-W and N-S resolutions of its smallest units (cells). Most of the raster and displays modules are affected by the current region settings. https://grass.osgeo.org/grass70/manuals/g.region Define the current region specifying coordinates and resolutions
  • 45. 45 GEOlab, Politecnico di Milano – Como Campus GRASS region g.region module manages the boundary definitions for the geographic region. The region is a geographic area with defined boundaries (N,W,S,E) and specific E-W and N-S resolutions of its smallest units (cells). Most of the raster and displays modules are affected by the current region settings. Define the current region settings to match the extension of a raster/vector map https://grass.osgeo.org/grass70/manuals/g.region
  • 46. 46 GEOlab, Politecnico di Milano – Como Campus Re-projection: v.proj / r.proj v.proj / r.proj modules re-projects a vector / raster map from one location to the current location. 1. Create a location in the layer reference system 2. Import the layer into the location 3. Create the location in the reference system you want to re-project the layer into 4. Run v.proj (for vector map) or r.proj (for raster map) modules https://grass.osgeo.org/grass70/manuals/v.proj.html https://grass.osgeo.org/grass70/manuals/r.proj.html
  • 47. 47 GEOlab, Politecnico di Milano – Como Campus Merging: r.patch r.patch module creates a composite raster map layer by using known category values from one (or more) map layer(s) to fill in areas of "no data" in another map layer. https://grass.osgeo.org/grass70/manuals/r.patch.html
  • 48. 48 GEOlab, Politecnico di Milano – Como Campus Rasterization: v.to.rast v.to.rast converts a vector map into a raster map. The resolution of the output raster map is the same of the current region. https://grass.osgeo.org/grass70/manuals/v.to.rast.html
  • 49. 49 GEOlab, Politecnico di Milano – Como Campus Reclassification: r.reclass r.reclass reclassifies a raster map based on category values. https://grass.osgeo.org/grass70/manuals/r.reclass.html
  • 50. 50 GEOlab, Politecnico di Milano – Como Campus Matrix confusion calculation: r.kappa r.kappa calculates error matrix for accuracy assessment of classification result. https://grass.osgeo.org/grass70/manuals/r.kappa.html
  • 51. 51 GEOlab, Politecnico di Milano – Como Campus Matrix confusion calculation: r.kappa r.kappa calculates error matrix for accuracy assessment of classification result. https://grass.osgeo.org/grass70/manuals/r.kappa.html Stats.txt Confusion matrix Commission and omission errors Overall accuracy