Mapping and monitoring of agricultural drought across different land uses and land cover in the North-Eastern KwaZulu Natal

Abstract

Drought is complex and one of the least understood natural hazards in Southern Africa. Timely information about the extent, the intensity, duration and impacts of the agricultural drought is essential for adaptation and management. In this study, the research aims, are made to monitor and map agricultural drought across different land uses and land cover in north-eastern KwaZulu-Natal as it was declared a disaster area in 2016 (AgriSA, 2016). Droughts occurred throughout South Africa during the summer season of 2014 to 2015 and 2015 to 2016. In this study the adopted methodology was through the use of remote sensing and Geographic Information System (GIS) techniques. Remote sensing and GIS was used to map and monitor the agricultural drought in the study area. To understand the impacts of the drought across different agricultural land use and other land cover types, the land uses and land cover was classified using Landsat earth observation data and maximum likelihood algorithm in the study area, and multi-temporal Normalized Difference Vegetation Index (NDVI) (1997-2017) with a twenty year interval used to map and monitor the agricultural drought and the meteorological (rainfall) in order to validate the NDVIs. Agricultural drought was then determined from investigating changes between 2015 and 2017 which were years that experienced severe conditions. The rainfall data was interpolated using Inverse Distance Weighted (IDW) interpolation to understand the mean rainfall from the weather stations services. Thereafter, Standardized Precipitation Index (SPI) values were determined from the rainfall data in order to understand the severity of the

droughts in certain parts of the study area from the weather station data. The meteorological analysis was cross compared with agricultural drought. The mean NDVI and mean rainfall interpolated shows that their relationship is inversely proportional, because where rainfall is low; NDVI is high for the years 2015 to 2017. The land use and land cover in the study is largely dominated by bush, cultivated cane crop, grassland and plantations. Looking at the overall classification in the year 2015, it is clear that bush land use and land cover was largely dominated in the study area, with other land use and land cover classes which were also part of the year 2015. During the year 2016 the other classes of land use and land cover where also dominating the study area for example grasslands and plantations. In the year 2017 we see cultivated cane crop start to emerge in the study area but land use and land cover is largely dominated by bush land use and land cover. The overall accuracy of the study was 74.2%.

Keywords: Agricultural drought, Land use/land cover, Remote sensing, Landsat 8 OLI/TIRS, Normalized Difference Vegetation Index, Standardized Precipitation Index, Accuracy Assessment.