An Evaluation of 20-m ESA-CCI S2 Prototype LC Product

Adoum Alkhalil,Michel Kouadio,Ibrahima Kadaoure

doi:10.3389/fsufs.2020.504334

Adoum Alkhalil, Michel Kouadio + Show 1 more

Open Access

https://doi.org/10.3389/fsufs.2020.504334

Copy DOI

Abstract

Agricultural production represents a key element in food security analysis. However, precious early warning time is often lost before the agricultural statistics are processed and released. The objective of the current study is to assess the performance of the 20-m ESA-CCI (European Space Agency - Climate Change Initiative) S2 prototype LC map classification in comparing it with Sentinel 2A 10 m or Google Earth images. In case the results turn out satisfactory the cropland class could be use as a crop mask to estimate crop production. The assessment method consists of 1) selecting six equal size polygons so that the Sahelian and Sudanian zones of West Africa are represented at each one of the western, central and eastern basins of the Sahel 2) generating four hundred dots randomly overlaid on the sample polygon 3) making extractions at dot location from the map under assessment and using Sentinel 2A 10 m or Google Earth to identify and count through visual interpretation the dots that fall over cropland and the ones that fall on other classes. The last two steps are repeated 10 times to bring the size of the validation point sample to four thousand for each of the 6 sample polygons. Our major concern being the assessment of the cropland classification, the number of classes is reduced to two (cropland and other) with the precision and the recall used as performance indicators. Analysis of the precision indicates that the classification of cropland in the Sahelian zone is less than 3% correct in East Sahel, less than 7% in West Sahel and about a third of the time correct for the Center Sahel. This is easily explained by the fact sand dunes and degraded land that makes up a significant part of the area have been mistakenly taken for cropland. Even in the Sudanian zone where the classification performance is better, the highest precision indicates that cropland classification is incorrect for a little over 28% of cases. Therefore, the ESA-CCI S2 prototype land cover map can’t be used as a crop mask.

Highlights

At a global scale changes in Land Use and Land Cover close monitoring is extremely important considering their intricate relationship with global change and global warming
They are lowest for East Sahel and West Sahel with 0.3939 and 0.4405, respectively. This basically means the ESA-CCI S2 LC classification performs better in the Sudanian zone than it does in the Sahelian zone
Because of the disproportionate size of the class “other,” accuracy is not a good classification performance indicator; it is necessary to continue the analysis with other indicators

Summary

Introduction

At a global scale changes in Land Use and Land Cover close monitoring is extremely important considering their intricate relationship with global change and global warming. West Africa’s population accounts for 30% of Africa’s population and has increased by about five-folds from 1950 to 2015 and is expected to increase from a little above 367 million to about 570 million by 2035 Such a rapid demographic growth has been partially the cause of poverty and lack of economic opportunities, which has in turn resulted in an increased pressure on the natural resources. A recent Land Use-Land Cover collaborative study between the Inter State Committee for Drought Control CILSS1 and USGS/EROS2 (CILSS, 2016) revealed that agricultural expansion accounts for most of the changes that occurred from 1975 to 2013 During this period cropland doubled to reach a total of 1,100,000 km or 22.4% of the land surface. With such drastic agricultural land expansion fallow technique is practically overlooked; cropland and cropped area are almost synonymous

Objectives

Results

Conclusion