Drought impact assessment in Southern African biomes

Abstract

This thesis investigates the effects of meteorological drought on Southern African biomes using remotely sensed rainfall and vegetation data. To achieve this objective, a review of studies on meteorological drought impact on vegetation was first provided. Secondly, we analysed pixel-wise spatio-temporal rainfall trends based on the space-time 3D rainfall cube approach and then assessed the meteorological drought impact on biomes by using Vegetation Condition Index (VCI) data. The vegetation response to drought impact (i.e. the time lag of vegetation response to drought impact) was calculated by correlating the Standardised Precipitation Evapotranspiration Index (SPEI) and the Normalised Difference Vegetation Index (NDVI) at different SPEI time scales. Finally, we analysed drought impact in relation to land use and land cover using land cover data (Globcover 2009) and a novel land cover layer “village pixel” developed based on the livestock density data. The village pixel is a proxy for rural communities and provides a rare opportunity to analyse drought impact on rural communities who mostly depend on livestock and rainfed agriculture. The novelty of this study is its regionality (i.e. covering all the Southern African biomes) and the fact that all the computations were done at pixel level.

The results of the Mann-Kendall trend analysis revealed a significant (P<0.05) decrease of rainfall over most parts of the biomes mainly over the Forest biome and south western parts of the region (i.e. Fynbos, Desert and Karoo biomes).An increasing rainfall trend was mainly observed over the central parts of the Southern African region mainly covering the central and western parts of the savanna biome. In assessing the drought impact on vegetation, the results showed that extreme drought impact on vegetation is mainly confined to the south western biomes. We also observed an increasing trend of seasonal VCI (1998-2017) across all biomes implying a decrease in the intensity of drought impact on vegetation. This result is of interest considering the general declining rainfall over most parts of the Southern African biomes. A plausible hypothesis to explain this is that the proportions of tree cover have been increasing over time. This phenomenon is referred to as woody encroachment.

The time lag of drought impact on vegetation (i.e. vegetation response to drought impact) results showed that the biomes do not immediately respond to drought impact. The 1 to 3-month SPEI time-scale, which shows the water balance for the current and previous two months had the highest correlation between SPEI and NDVI over most parts of the arid biomes (i.e. Savanna, Grassland, Nama Karoo) and humid biome (i.e. Forest). On the contrary, the Desert, Succulent Karoo, Fynbos and Montane biome responded to drought at medium to long time scale (i.e. 6 to 24 months). The assessment of drought impact in relation to land cover and land use based on the Kruskal-Wallis test showed a significant difference (p<0.05) in drought impact among the different land cover types. This confirmed the fact that drought impact of vegetation varies with land cover types. It is worth noting despite the fact that the evergreen forest and flooded vegetation recorded the lowest seasonal mean VCI during the 2015-2016 drought, the lowest dekadal VCI values were recorded within the rural areas “village pixels” land cover, indicating the vulnerability of rural communities to drought impact. The analysis also revealed that shrubland and grassland were least affected by drought.

Whilst future drought cannot be prevented, the results of this study provides information that can be used to ensure that communities are better prepared to cope with future droughts. The results from the analysis of rainfall trends can help to map drought hot spots.