Colloquium

Remote sensing and crop recognition: improving the information system around smallholders in Mali

Organisator Laboratory of Geo-information Science and Remote Sensing
Datum

di 5 april 2016 10:00 tot 10:30

Locatie Gaia, building number 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 2

By Wilmar van Ommerren (the Netherlands)

Summary

Creating a transparent information system is crucial to improve the position of smallholder farmers in Mali. Therefore, the project Spurring a Transformation for Agriculture through Remote Sensing was launched and one of the main goals of this project was crop type classification. Crop variability, crop similarity at certain growing stages, different land preparation practices and landscape characteristics were, however, posing challenges for a successful classification. Several classifiers, indices and strata were tested on their suitability for the classification of two study sites. Therefore, both UAV and satellite imagery was made available. Classifications were carried out on a single image and on multi-temporal datasets. In addition, the influence of the spatial and temporal resolution of the images was tested. It was discovered that when crops grew under optimal conditions, a classification with a single image, captured while the crops were still flowering, was sufficient (overall accuracy of more than 80 percent). For a more heterogeneous study site a stratified temporal classification was more suitable. Increasing the spatial resolution was only useful if the training dataset contained a relatively large number of extreme observations. However, increasing the temporal resolution was found to be useful in all classifications. A reduction from eight to five images did not have a large influence on the accuracy. The maximum overall accuracy for the heterogeneous study area was 68 percent when an extensive training dataset covering the entire study area was used. Crop classification on smallholder farms in Mali is thus possible. Moreover, the approach followed by this research is flexible and can be applied to other study areas.

Keywords: pixel-based classification; smallholder farms; Mali, multi-sensor; spatial resolution; temporal resolution; per-field aggregation; vegetation indices; multiple classifiers; mono-temporal classification; multi-temporal classification.