University of Twente Student Theses

Login

Remote Sensing-Based Characterization of Biodiversity Supporting Structures on Coffee Farms in Zimbabwe

Chimbi, A.A. (2024) Remote Sensing-Based Characterization of Biodiversity Supporting Structures on Coffee Farms in Zimbabwe.

	PDF 2MB
	PDF 27kB

Abstract:	Coffee is an important commodity crop in international trade and mainly produced in tropical areas where deforestation is reported to be happening. There is a growing market demand for sustainable coffee that is deforestation free and supports ecosystem services among other social and environmental requirements. Current methods of assessing indicators of sustainability rely on sending assessors and this is expensive, time consuming and often subjective. The goal of this project was therefore to develop an accurate remote-sensing based method to map and quantify biodiversity supporting structures (BSS) on coffee farms to support coffee certification. Field data on the location, extend and other features of five BSS (hedgerows, shade trees, coffee, forests, and wetlands) were collected at two farms (Crake Valley and Jersey) in Zimbabwe. A random-forest based classification routine for optical data (Pléiades and GeoEye-1) and radar (Sentinel-1) data was implemented using the field data collected using spectral bands and vegetation indices. The separability of BSS using bands, indices and radar metrices was tested then the best performing indices were used for the mapping and quantification of the BSS from the remote sensing data. Using the mapped BSS, a change assessment was then implemented to assess if there are changes in the BSS at the two farms. Using the Fishers linear discriminant analysis on optical data, three vegetation indices (NDVI, NDWI and GNDVI) at Crake Valley ranked the most discriminant while SAVI, EVI and B3 were the most discriminant at Jersey. Different Sentinel-1 radar based data types were tested to see if using them in mapping BSS would improve the accuracy but it was concluded that this approach was not successful because the low spatial resolution to recognize small BSS. Applying the best discriminating features to map the BSS at Crake Valley produced an overall accuracy of 73.1% and 52.0% for optical and radar respectively. At Jersey, the overall accuracy was 64.9% and 38.5% for optical and radar respectively, with coffee and forest being difficult to distinguish. It was concluded from this that the optical methods are better in mapping BSS in coffee farms and were applied to detect if there were changes in these at the two farms. Crake Valley had higher density of BSS in relation to coffee compared to Jersey meaning increased ecosystem services. It is concluded from this study that it is possible to map and quantify BSS on coffee farms with high resolution optical remote sensing data to support sustainability assessments with an accuracy above 64%. Further work is required to standardize the methods across farms and to include other indicators in the assessment.
Item Type:	Essay (Master)
Faculty:	ITC: Faculty of Geo-information Science and Earth Observation
Subject:	43 environmental science
Programme:	Geoinformation Science and Earth Observation MSc (75014)
Link to this item:	https://purl.utwente.nl/essays/101952
Export this item as:	BibTeX EndNote HTML Citation Reference Manager

Show download statistics for this publication

Repository Staff Only: item control page