Mapping rice and rice growing
environments in West-Africa using
remote sensing and spatial
modelling tools
Sander Zwart
1. Importance of rice production for food security in West-Africa
2. Rice production environments
3. Strategy for rice mapping in West-Africa 4. First results
Rice and food security in West-Africa
Rice production and consumption in Africa (1970-2010)
Rice and food security in West-Africa
Contribution of various staple crops in diets in West-Africa (1961-2010)
Rice and food security in West-Africa
Challenges for food security:
(West-)Africa is by far not self-sufficient and depends on international markets.
Climate change is impacting W-Africa strongly;
less rainfall and more erratic and intense rainfall events, lower river discharges, floods.
Rice and food security in West-Africa
Why do we need creating maps and statistics of rice?
• (sub-)national rice statistics are very unreliable or absent in Africa
• Understanding where rice is for efficient
targeting of technologies, interventions and actions
Rice production environments Rainfed upland and lowland Smallholder fields Intercropping Fragmented landscape Inland valleys / wetlands Very dynamic
One rice crop
Irrigated rice Large-scale systems Gradual expansion Two seasons Mangrove rice
Cleared lands in forested areas Stable systems
Rice production environments
Differences between Asian and African rice landscapes
Asia Africa
Irrigated rice (80%) upland rainfed lowland rainfed
lowland irrigated (~10%) Stable area Dynamic & expanding 30% of arable land 4% of arable land
Contiguous rice areas Fragmented
Paddy land preparation Dry land preparation High fertilizer inputs Low fertilizer inputs
Strategy for rice mapping in West-Africa Rainfed upland and lowland Radar RS Spatial modelling Random Forest Irrigated rice Optical RS Radar RS Mangrove rice off-season Landsat GoogleEarth interpretation
Strategy for rice mapping in West-Africa
An assessment of the rice growing areas in planned using data no older than 5 years.
Irrigated rice: Landsat 8 imagery, supervised
classification
Use of radar imagery planned
Mangrove rice: Landsat 8 imagery (off-season)
GoogleEarth interpretation
Rainfed systems: spatial modelling
Random Forest Radar imagery
First results – irrigated rice
Pilot testing of radar remote sensing in two hubs:
Cosmo-SkyMed imagery is acquired every 16
days during rice season Spatial resolution of 3m
Senegal: irrigated rice conditions (July-December) Benin: upland and lowland rice (June-december)
Goals: mapping rice and assessing crop phenology dates (SoS and harvest)
First results – irrigated rice
First results – mangrove rice
1. Visual interpretation and digitization in GoogleEarth (2010-2014 high resolution satellite images)
2. Remove water and mangrove forest patches using off-season NDVI maps derived from
Landsat 8 imagery
Implemented in Senegal, The Gambia,
Guinea-Bissau, Guinea-Conakry, Sierra Leone, Liberia Total of 11 Landsat scenes
First results – mapping inland valleys / wetlands
Inland valley or wetlands (irrigated and rainfed
lowland)
• Areas suitable for rice production due to favorable hydrological conditions
First results – mapping inland valleys / wetlands
stream
20 20 21 21 23 23 24 25m 25 24 altitude (m) 30mDigital Elevation Model
(2-dimensional)
Selected inland valley bottom
First results – mapping inland valleys / wetlands
Validation
Omission/comission errors, accuracy and area estimation/comparison
First results – mapping inland valleys / wetlands
Currently only 10% cultivated (official stats)
Mapping rice in the inland valleys using remote sensing classification is (currently) impossible: • valley size
• heterogenous agricultural landscape • image resolution
• extent
First results – mapping inland valleys / wetlands
Random Forest
Machine learning technique based on the
construction of decision trees that can be used for regression or classification purposes
Predict the presence of rice cultivation in the inland
First results – mapping inland valleys / wetlands
• Collection of data on inland valleys and
presence or non-presence of rice or agriculture • Building geo-spatial data bases containing:
Road networks, villages, travel distance, markets (inputs and outputs), population density, inland
valleys, soil types, water availability, rainfall (remote sensing), etc.
• Implementation in Sierra Leone, Liberia, Benin and Mali
Challenges for rice mapping
• Rainfed upland agriculture might be too
fragmented, too small scall-scale, too dynamic, to be able to identify.
• Skilled person-power
- Very few young people are educated in GIS/RS
- (Almost) no experience with radar remote sensing.