Crop identification and acreage estimation: Improve satellite recognition of dominant crop species (cotton, maize, millet, sorghum, peanut) in smallholder fields, using VHR optical satellite images as well as C-Band and X-Band SAR data as input.
Residue and Tillage mapping: n/a
Soil moisture: n/a
Crop biophysical variables (LAI): i/ Monitor canopy development states for dominant crops including fraction vegetation cover, planting density, LAI, canopy height and density. ii/ Improve seasonal yield and biomass prediction for crops and forage. The focus here will also be on microwave data as input.
Other: n/a
The mapping resolution is sub-meter to 5-m. The timeliness (with regards to growing season) is 01 April- 30 November. Research activities on crop identification and acreage estimation will be linked to those on crop biophysical variables characterization.
Project Reports
Implementation Plans
1. Crop identification and acreage estimation:
2014: Algorithms development and calibration in Kani-Sukumba (Mali)
Land use / land cover survey for 150 smallholder fields (30 fields for each of the 5 dominant crops, each field at least 1ha in size)
Compilation of historical (2002-2013) land use land cover data into consolidated database (depending on the year of survey, data covers between 150-500 fields)
Acquisition of archive TerraSAR-X and Radarsat2 data (if available) and tasking of new data acquisitions, with a target of at least 5 acquisitions for the season (field preparation/pre-sowing stage, vegetative stage, flowering stage, grain-filling stage, post-harvest). 10-day frequency preferred.
Algorithm development in partnership with competent ARIs (to be determined)
2015: Algorithms validation and crop area mapping in Sukumba, Kani, Nanposela (Mali)
Same as 2014, plus:
Forward application of algorithms developed in 2014 cropping season in Sukumba (same geography, different climate) including algorithm re-calibration
Forward application of algorithms developed in 2014 cropping season in Kani, Nanposela (different geography, different climate)
Peer-reviewed publications on libraries and methods
2. Residue and Tillage mapping: n/a
3. Soil moisture: n/a
4. Crop biophysical variables:
2014: Development of spectral, temporal and contextual (textural) libraries in Sukumba (Mali)
Detailed ground characterization of field-scale agronomic practices, environmental conditions (climate, soil), monitoring of canopy growth and collection of end-of-season yield and biomass data for 50 smallholder fields (10 fields for each of the 5 dominant crops, each field at least 1ha in size)
Tasking of TerraSAR-X and Radarsat2 data, with a target of at least 5 acquisitions for the season (field preparation/pre-sowing stage, vegetative stage, flowering stage, grain-filling stage, post-harvest). 10-day frequency preferred.
Concurrent collection of decadal (10-day) multispectral (VIR/NIR) data from WorldView2/3 and UAV (both NIR camera on-board of fix wings UAV and TetraCaM on-board of octocopter vehicle) under BMGF Remote Sensing Learning Package grant.
2015: field-to-landscape scale yield and biomass prediction for Sukumba (Mali)
Same as 2014, plus:
Calibration of appropriate yield and biomass prediction model for local conditions
Testing of multi-source satellite data assimilation for improved end-of-season yield and biomass predictions, in partnership with Univ. Catholique de Louvain
5. Other: n/a
The current project phase is research. The proposed project builds on 2002-2013 research investments in the locality of Sukumba under projects ‘Carbon From Communities’ (NASA, 2002-2004), ‘Soil Management CRSP’ (USAID, 2002-2007), ‘Seeing Is Believing – West Africa’ (AgCommons/BMGF, 2009-2010) and Dryland Systems CRP (CGIAR, 2013-present). It also leverages the upcoming ‘Imagery for Smallholders – Activating Business Entry points and Leveraging Agriculture’ (ISABELA) project (BMGF, 2014-2015).
Site Description
Locations
Mali - Nanposela
Site Extent
Centroid:
12.329, -5.3225
Top left:
12.392, -5.387
Bottom Right:
12.266, -5.258
Mali - Kani-Sukumba
Site Extent
Centroid:
12.176142, -5.189662
Top left:
12.220957, -5.235968
Bottom Right:
12.13132, -5.143373
Typical field size range: 1.4±1.2 ha.
Dominant crop types are:
cotton (Gossypium hirsutum L.) – typically 40% of planted area
maize (Zea mays L.) – 15% of planted area
pearl millet (Pennisetum glaucum (L.) R.Br.) – 20% of planted area
sorghum (Sorghum bicolor L. Moench) – 20% of planted area
peanut (Arachis hypogaea L.) – 5% of planted area
Typical field rotations are:
cotton – millet/sorghum
cotton – maize – millet/sorghum
cotton continuous
However, although these are the most common rotations observed, deviation from these is the norm rather than the exception.
The typical cropping calendar is:
cotton: field preparation Mar-Apr, sowing May-Jun, harvest Oct-Nov
maize: field preparation Apr-May, sowing Jun-Jul, harvest Sep-Oct
millet-sorghum: field preparation Apr-May, sowing Jun, harvest Nov-Dec
peanut: field preparation Apr-May, sowing Jun-Jul, harvest Oct
Sowings vary significantly inter-annually depending on the onset date of rains. Continental West Africa is among the regions of the world where onset date of rains is most variable and unpredictable. Wild and domesticated plants have evolved traits to evade the associated risk, such as photoperiod sensitivity.
The climatic zone is ‘tropical wet and dry or savannah’ (Aw in Koppen’s classification). Climate (in nearby Koutiala, 1971-2000) is subhumid with annual rainfall ranging from 704 mm for a dry year (probability of exceedance =0.9) to 838 mm for a median year (=0.5) and to 1,058 mm for a wet year (=0.1).
Soil Texture
Soils include Haplustalfs (PIRT, 1983 local soil classification: PL9, PS3), Haplustults (PL11), Cuirorthents (TC5), and Cuirustalfs (TC4) with shallow slopes, locally significant rock outcrops, and few erosion gullies. Sandy loams dominate. More complete corresponding soil profile information is provided in annex 1 in DSSAT format.
Landscape Topology
Landforms, distribution of soils and cropping systems is analogous to that reported by van Staveren and Stoop (1985):
Figure 1a: schematic representation of soil conditions, water movement and cropping systems along a comparable toposequence in Kamboinse, Burkina Faso (van Staveren and Stoop, 1985)
Figure 1b: schematic representation of relationships between staple crops, sowing dates, and land types for a comparable toposequence in Burkina Faso (van Staveren and Stoop, 1985)
Soil Drainage Class
Most soils (PL11, PS3, TC4) are well drained. Some are imperfectly (PL9) or somewhat excessively (TC5) drained.
Other Site Specifications
Irrigation infrastructure is absent. Lowlands are used for high value crop (vegetables, orchards) and locally rice production. A map of the project site is provided in annex 2.
Specific Project Objectives & Deliverables
Specific Project Objectives and Deliverables
Mapping Agricultural Areas:
Land Cover: Mapping Frequency: Yearly, preferably around peak biomass (September).
Cropped Land: Mapping Frequency: As above.
Pasture/Rangeland: Mapping Frequency: As above.
Estimating Crop Areas
Statistical Units: field scale (ha)
Crop Growing Conditions Over the Growing Season (for 50 fields):
Daily tmin, tmax, precipitation, relative humidity, solar radiation, wind speed/direction (1 automatic weather station installed at site)
Vegetation reflectance (VIS/NIR) from UAV, LAI, phenology (on the ground, in-situ measurements) on a 10-day frequency
Non-destructive and destructive biomass measurements on a 10-day frequency
Estimation of Biophysical Variables (from satellite):
NDVI, LAI, fCover, fAPAR, biomass on a 10-day frequency
Phenological Events (from in-situ measurements):
Field preparation, sowing, emergence, juvenile, vegetative, flowering, grain filling, maturity, harvest, post-harvest dates / durations will be documented for each of the 50 fields monitored throughout the cropping season (see in-situ observations section below)
Information about crop development stresses (abiotic, biotic) will also be collected as part of the same field monitoring protocol
Integration of the EO-Derived Information Into Crop/Agro-Met Models / Forecasting Agricultural Variables from Crop/Agro-Met Models:
Data assimilation of LAI estimates will be explored in 2015 in hindcast (2014 data) and forecast (2015 data) modes. The crop/agro-met model to be used will likely be AQUACROP (subject to confirmation – other options are possible using the already locally calibrated, but more complex SarraH, APSIM and DSSAT models).