Ramesh Gautam, Jean Woods, Simon Eching, Mohammad Mostafavi Land Use Section, Division of Statewide...
-
Upload
avis-henry -
Category
Documents
-
view
216 -
download
0
Transcript of Ramesh Gautam, Jean Woods, Simon Eching, Mohammad Mostafavi Land Use Section, Division of Statewide...
Crop Mapping in Stanislaus County using GIS and Remote Sensing
Ramesh Gautam, Jean Woods, Simon Eching, Mohammad Mostafavi
Land Use Section, Division of Statewide Integrated Water Management
California Department of Water Resources
Usefulness of Land Use Mapping
Quantify crop acreage based on crop types
Estimate evapotranspiration
Determine urban landscape acreage
Input for groundwater and surface water
models
Verify fields fallowed for water transfers
Map urban growth patterns
Estimate economic impacts of flooding
Why Remote Sensing Based Crop Mapping is Needed
Reduce the extent of required field mapping by identifying permanent crops
Accurately assess crop acreage
Estimate annual crop water use for the California
Water Plan
Accurately estimate evapotranspiration on a field
level
Determine the annual extent of fallowing
Verify fields fallowed for water transfers
Study Area
Stanislaus CountyArea: 1,515 sq mile
Population: 515,000
Overall Crop Mapping Strategy
All Crops
Decision Tree Based
ClassificationOrchards
Non-Orchards
LCRAS Based
Classification
Time series based
Vegetation Index
Analysis
Corn, Mixed Pasture,
Fallow, Dry Beans, Tomato, Melons
AlfalfaAutocorrelation & LIDAR
Vineyards
Classify orchards from non-orchard crops
Gray Level Co-occurrence Matrix Algorithm was used to classify orchards
Textural patterns distinguish orchards from other crops
eCognition Developer software was used to develop the algorithm
Decision Tree Classification Technique
Data Processing
Textural parameters are analyzed to evaluate the fields having coarse texture versus fine texture
First Level of Classification: Results
Recently planted orchards were classified in next level as shown in next slide
Non-orchards
Orchards
Bare land and new orchardsFarmsteads
Urban area
Poultry farms
Highways/Roads
LEGEND
How recently planted orchards have been captured in second level of classification
Non-orchards
Orchards
Bare land and new orchardsFarmsteads
Urban area
Poultry farms
Highways/Roads
LEGEND
Second Level of Classification: Results
Non-orchards
Orchards
Bare land and new orchardsFarmsteads
Urban area
Poultry farms
Highways/Roads
LEGEND
Final Classification
Mapping Orchards in Stanislaus County
Mapping Non-Orchards using Lower Colorado River Accounting
System (LCRAS)
Ground Truth Survey
Collect Crop Attributes(12% of Total Fields)
QC Ground Truth Data
Update Field Border Database
Develop Personal Geo-database of Ground-truth data in ArcGIS
Randomly Select Training Data (60%)
Perform Image Segmentation in eCognition Developer
for Training Data
Create Signatures in Erdas Imagine
Data Processing
Using eCognition Developer software, crop fields are segmented into polygons of similar spectral characteristics.
LCRAS Classification Method Cont’d…
LANDSAT-5 ImageBands 1-5 and 7
Perform Supervised Classification of Spectral
Characteristics
Identify Crops at the Field Level Based on Classification
Perform Accuracy Assessment
Re-evaluate signature sets
Identify Mislabeled Fields Based on
Ground Truth
Overall Classification ≥ 90%?
Yes
EndNo
Year 2010 Crop Map, Stanislaus County, California
Staff Time Requirements for Crop Classification
Ground Truth Survey15%
Decision Tree Based Classification of Permanent Crops
25%
LCRAS Based Classi-fication of Annual
Crops60%
Questions?