Cropland and tree cover mapping using Sentinel-2 data in an agroforestry landscape, Burkina Faso

Detta är en Master-uppsats från Göteborgs universitet/Institutionen för geovetenskaper

Sammanfattning: Sentinel-2, with high spatial resolution bands and increased number of spectral channels,has provided increased capabilities for vegetation mapping. Cropland masks withinheterogeneous areas such as the Sudano-Sahel zone have become useful for monitoringlandscapes. The objectives of this study were to assess the utility of Sentinel-2 data inclassification of cropland for the purpose of creating a cropland mask, and estimation oftree cover. An assessment of the cloud-free, wet season satellite images from 2017 and2018 (15 in total), from the Saponé agroforestry parkland landscape in Burkina Faso wasconducted. The random forest machine learning algorithm is applied to images to performclassification with field-based data as training data, tree crown cover estimation with highresolution Pléiades image and to assess variable importance. The results reveal that dueto the dynamic cropping practices, the cropland mask needed to be produced for a singleyear at a time, and high model accuracy was indicated for 2017 with overall accuracy of94.7%, yet lower for 2018 (90.9%), even though similar acquisition image dates were used.The best result for 2017 was produced using multi-temporal images from October 7 and22, while the best result for 2018 was obtained using a single image from October 22.Variable importance measures revealed that the green, NNIR, red, NIR and vegetationred edge5 bands were most important in both 2017 and 2018 analysis. The percent of treecrown cover was estimated for 2017 using Sentinel-2 images from June 29 and October22 and a random forest regression algorithm. The R2 of the best regression equation was0.42 with a RMSE of 15.1. The RF prediction had values ranging from 0.52% to 85% treecover. The relationship between observed and predicted tree cover was linear, however,there was an underestimation of higher percentage tree cover values and an overestimationof very sparse tree cover. Based on the results, Sentinel-2 may be useful for monitoringcropland at landscape level and identifying tree crown cover. However, this study wouldhave benefited from using more discriminating field-based training data (i.e.crop typesand harvested fields) to identify active cropland. In conclusion, the Sentinel-2 data, withits 10 m pixels and range of spectral bands in particular the red and vegetation red edgeproduced good quality cropland masks. The use of high resolution supplementary image(Pléiades) is also recommended as a source of training data for producing cropland masksand tree cover data. The results presented here will contribute to an ongoing researchproject on the role of trees on agroforestry landscape productivity.

  HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)