Detecting changes in cropland extent: A case study within the Lake Chad Basin using classified remote sensing data

Abstract: Access to fresh water is essential for the lives of organisms and for human activities. However, with a growing population that strives after a higher standard of living, the pressure on already limited resources will rise. Lake Chad is one out of many diminishing lakes in the world where the call for action is urgent to prevent ecosystem degradation, famine, and outbreak of war. The lake has shrunk by more than 90% since the 1960's due to decreasing levels of precipitation and changes in water consumption. As most dwellers in the Lake Chad Basin are farmers and the agricultural sector consumes the largest proportion of all water resources, change in irrigation patterns is a likely cause for this decrease. Analysis of the change in cropland extent using classified remote sensing data acquired between 1992 and 2009 could possibly delineate areas where this change has occurred. The result could later be used to guide further studies on where suitable measures should be implemented to stop the current degradation. Thus the main objective of this thesis is to delineate areas of change in cropland between 1992 and 2009 that could possibly explain the increased water consumption in the Lake Chad Basin. The study also aims to discuss the suitability and reliability of using classified land cover datasets to detect these changes. Classified land cover datasets are generated based on multispectral images that are acquired by monitoring sensors onboard satellites. The spectral characteristic of a cell reviles properties of the observed surface and this information can be used to classify each cell in an image into a unique land cover type. Three types of datasets are used in this analysis: the MCD12Q1, the Globcover and the GLCC. They are all based on data acquired during different time periods, from different sensors and with different spatial resolution. The datasets from 1992-1993, 2007 (GLCC and MCD12Q1 respectively) and the two Globcover datasets (2004-2006 and 2009) are found to not be suitable for this type of analysis. The MCD12Q1 datasets, on the other hand, were produced on an annual basis from 2001 to 2007 and give a clear indication of the expansion of cropland in certain areas in the basin. A constant intensification of cultivated land near Lake Chad’s northern pool and a migration of cropland from the upper KYB downstream the rivers could be detected. The flood and recession farming along the lower KY also seem to have increased between 2001 and 2007. Despite the convenience in using classified remote sensing datasets to detect changes in cropland extent within the Lake Chad Basin, there are several drawbacks. Unique conditions in every observation, spectral resolution, spatial resolution and classification method are some of the reasons why different land cover datasets are not compatible with each other. Local conditions in the Sahel, such as almost permanent cloud cover and mixed vegetation, limit the possibilities to acquire useful data from this region. The lack of classified land cover data from 1960 to 1980, when the largest retraction of the lake occurred, limits the possibilities of finding factors that can explain the decreased water resource. Studies using Landsat multispectral images should therefore be carried out to delineate areas where change in cropland extent has occurred since 1972. This could possibly be a better indicator of the reasons to the increased water consumption since the 60’s.