Senegal - Bambey

Specific Project Objectives & Deliverables


Millet and Groundnuts Identification with Optical and Radar Sensors

Comparing the three maps that we created using Pleiades data alone (2013/01/16; 2013/12/31, 2014/10/30), results show that the millet and groundnuts crops are better classified during the dry season (2013/01/16; 2013/12/31), with respectively 81.4 and 82 % of good ranking. Indeed, contrary to all expectations, in the wet season (2014/10/30), the millet is only classified well in 47% and 30 % for the groundnut with the Pleiades image acquired in October 2014 used alone.
On the other hand, the combination of the Pleiades image with the Landsat-8 time series in 2014 allowed to improve the classification (Random forest classification) with 80% of millet and groundnuts well classified and also to discriminate other crops which was technically impossible during the dry season (Figure 1). Indeed, images acquired during the wet season allowed classifying the Niébé with 68.48%, the sorghum with 97.71%, the fallows with 89.83 % and the pastureland with 79.35% accuracies.

Figure 1 Land Use Map derived from the Combination of one Pleiades Image and Landsat-8 Time Series Acquired during the Wet Season 2014


Using the Normalized Difference Vegetation Index (NDVI) of the Landsat-8 image series, we also studied the phenological state of the main cultures (see Figure 2 below). The NDVI allows one to follow the state of the cultures during the wet season. Most of them began to develop their leaves around August 15, and were mature with dense vegetation cover at the beginning of October 2014.

Figure 2 NDVI Temporal Profiles derived from Landsat-8 Time Series of the Main Cultures in the Bambey Area during the Wet Season 2014


Millet and Groundnuts Conditions with a Radar Sensor during the Rainy Season 2014

We also tested a time series of TerraSAR-X radar images in Bambey study area for crop discrimination and also to study the dynamics of their phenological states. After image calibration, different polarimetric parameters such as Shannon entropy and Pauli decomposition and methods as unsupervised classification based on H/A/α parameters and the SEaTH algorithm were performed. Unfortunately, these methods were not as effective in discriminating groundnuts from millets or in following their phenological states as the optical sensors tested (Figure 3).


Figure 3 Millet and Groundnuts Temporal Profiles Showing the Shannon Entropy Mean derived from TerraSAR-X Time Series in Bambey Area during the Wet Season 2014


Evolution of the Crops and Trees since 1968

Our aim in this study is to follow the agricultural spatiotemporal dynamics in the Niakhar area close to Bambey village since 1968. We want to measure the evolution of the crop to the detriment of the natural vegetation and more especially the trees which are very important for soil fertilization, for biodiversity conservation and also for the people’s needs (firewood, medicine, fodder…). To achieve that, we acquired from the CSE Institute archived satellite photography from the Corona American mission in 1968, which had 2 m spatial resolution. We also acquired two Spot 6-7 satellite images within the framework of the JECAM Program. We recently performed the ortho-rectification of the images and now we plan to map trees (species if possible) and the groundnut and millet crops.

Figure 4 Corona Images, 1968

©2015 Joint Experiment for Crop Assessment and Monitoring