the use of spot imagery in support of crop area estimates in south africa by geoterraimage | spot...
DESCRIPTION
Agriculture and satellite imagery: the use of Spot imagery in support of crop area estimates in South AfricaTRANSCRIPT
Spot Image and its partners add value to
satellite imagery in Agriculture
The use of Spot imagery
in support of crop area estimates
in South Africa Fanie FERREIRA
An applicative session on
The use of Spot imagery in support of crop area estimates in
South Africa
Fanie FERREIRA
GeoTerraImage
National Crop Statistics Consortium
• Agricultural Research Council– Institute Soil Climate &Water:
• Yield modelling research
– Summer Grain Institute: objective yield – maize• Field measurements:
– Small Grain Institute: objective yield – wheat• Field measurements
• SiQ– Aerial surveys & telephonic interviews – Statistical processing
• GeoTerraImage– Satellite image processing– Crop type classifications
Overview
• Field Boundary Mapping– Stratification of Agricultural activity
• PICES: Producer Independent Crop Estimate Survey– Aerial Surveys: crop type area per province– Verified during the Gauteng census project:
• difference < 1.8%
• Crop Calendar– Understanding crop evolution/phenology
• Crop Type Classification– Image processing & classification of satellite imagery
• Classified Field Boundaries– Various applications
Use of satellite imagerySPOT4 / LANDSAT
Previous seasons 2006/7/8
Area calculation @ field level
Field crop boundary
In-season 2009
Satellite Analysis
PICES survey @ provincial level
SPOT5
Stratification
• Rate of interview refusal increased
• Requirement: Develop new methodologyProducer Independent Crop Estimate SurveyStratification: Field Boundary on 2.5m Spot5
Full refusals
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
2002/2003 2003/2004 2004/2005 2005/2006 2006/2007
Field Boundary Mapping
• Digitising manual to ensure consistency
• Crop Field Boundaries: every cultivated field– Cultivation, Irrigation, Smallholdings– Orchards (Horti/Viticulture), Subsistence
• Irrigation– Centre pivots
Province Stratification MappedFields Reduction % ReductionFree State 10,794,982 3,712,625 7,082,357 65.61%North West 5,776,803 1,921,927 3,854,876 66.73%Mpumalanga 4,118,568 1,103,706 3,014,862 73.20%
Advantage of Field Boundaries: Survey Area (Ha) Reduced
SA coverage: 13 million ha
PICES: Sample selection
Shortest routing algorithmPoints to be surveyedGauteng Province
PICES Infrastructure
PICES: aerial survey
Additional points used for image training
Selected fields with identified crop types
• Vast improvement: survey efficiency
• Support image classification
• Statistical calculated of area
Season Selected Additional TotalWinter2006 1628 2500 4128Summer2007 1605 6211 7816Winter2007 1774 6112 7886Summer2008 1602 5568 7170
PICES Crop Type Points: Freestate Province
Province Selected Additional TotalMpumalanga 1048 1852 2900Freestate 1582 2618 4200NorthWest 1288 1812 3100Gauteng 319 230 549Total 4237 6512 10749
PICES Crop Type Points/Province: Summer 2006
Crop Type Development
• Seasons– Temperature: Summer vs Winter – Summer rainfall vs Winter rainfall
• Vegetative growth– Annual vs Perennial– Annual: germination, growth, senescence, harvest– Perennial: deciduous vs evergreen
• Cultivation Practices– Field preparation: fallow / bare soil– Planting dates
Crop Evolution: Multi Season
R/S Process Sequence
• Selecting cloud free images
• Ortho-rectification
• Mapping cloud areas
• Building image sets: optimal cloud free
• Indices: NDVI, Tasseled Cap, PCA, BSI
• Select bands for best discrimination
• Image Calibration: PICES Crop Types
• Classification: Supervised – user defined classes
• Field Boundaries (shp) populate: Zonal Majority
Freestate Province
SouthAfrica Freestate % of SAArea:ha 2,551,800 1,020,000 39.97%Yield(ton/ha) 2.79 2.80Production 7,125,000 2,855,000 40.07%
SouthAfrica Freestate % of SAArea:ha 2,799,000 1,170,000 41.80%Yield(ton/ha) 4.54 4.21Production 12,700,000 4,928,000 38.80%
SouthAfrica Freestate % of SAArea:ha 2,427,500 955,000 39.34%Yield(ton/ha) 4.74 4.53Production 11,513,950 4,323,750 37.55%
Maize 2009
Maize 2007
Maize 2008
Classification Procedure
– Erdas: Supervised classification• Maximium likelihood combined with Parallelepiped
– Calibration / Training• Generate signature file from PICES crop types
• Buffer field boundaries -60m: remove edge pixel
– Heterogenity within field cause confusion– Evaluate spectral parametres
• Select spectral bands from profile: bands vs crops
• Class confusion/conflict – Set Std Dev
– Field parcel (shp) populated: crop type – Zonal majority function & record majority fraction
– Irrigated maize vs rainfed maize: 12% higher fraction
Classification Analysis
Maize Dominant Area
CropType Count Area Ha Avg Ha % Area Count Area Ha Avg Ha % AreaFallowWeed 946 31306 33 5.31% 1494 58210 39 9.87%Groundnuts 546 20715 38 3.51% 551 21640 39 3.67%Maize 8699 349990 40 59.37% 7958 323826 41 54.93%MaizeWheatPivot 210 4290 20 0.73% 125 2759 22 0.47%Pasture 4167 93817 23 15.91% 4420 104479 24 17.72%SoyaBeans 4 217 54 0.04% 1 26 26 0.00%Sorghum 6 61 10 0.01% 0 0 0 0.00%Sunflower 598 20762 35 3.52% 1663 56631 34 9.61%Wheat 1813 66835 37 11.34% 895 24477 27 4.15%WinterGrazing 89 1509 17 0.26% 109 2460 23 0.42%Total 17078 589504 35 17216 594509 35
Cropping Season 2007 Cropping Season 2008
Wheat Dominant Area
CropType Count Area Ha Avg Ha %Area Count Area Ha Avg Ha %AreaFallowWeed 4133 71683 17 16.49% 3921 73450 19 16.90%Maize 6948 140457 20 32.32% 8430 164280 19 37.80%Pasture 7229 94960 13 21.85% 6992 96639 14 22.23%Sorghum 6 52 9 0.01% 91 946 10 0.22%SoyaBeans 1183 21736 18 5.00% 1059 18028 17 4.15%Sunflower 375 7030 19 1.62% 639 12002 19 2.76%Wheat 4210 90152 21 20.74% 3524 67004 19 15.42%WinterGrazing 634 8580 14 1.97% 126 1739 14 0.40%Total 24718 434650 18 24782 434088 18
Cropping Season 2007 Cropping Season 2008
Maize Comparison: 2007vs2008
Spatial Distribution Cultivated area
Crop type classification
District level comparison:
Maize area / district
SoyaBean Comparison: 2007vs2008Spatial Distribution Cultivated area
Crop type classification
District level comparison:
Soya area / district
Conclusion• Integrated Processing Chain
– Based on Spot 4 & 5 imagery
• Spot 5 imagery provided complete coverage– Field Boundaries: Improved stratification
• Large reduction in area to survey: Reduced costs (8X)• Accuracy increased
• Spot 4 imagery regular recordings– Complete cloud free coverage
• Calculation of district level area / field level• Similar zones can be calculated• Visualise cropping patterns & trends• Valuable for agro-industry planning
‘‘Need the right imagery solutionsfor your job?
Ask for Spot Image.’’
www.spotimage.com