Vegetation dynamic and drought: South African savanna case study.

Semiarid rangelands are one of Africa’s most complex and variable biomes. They are a mosaic of land uses, where extensive livestock is the main economic activity, and agriculture is also crucial. They are highly controlled by the availability of water, e.g., pasture and rainfed crop production. Although the vegetation is adapted to variable climatic conditions and dry periods, the increase in drought intensity, duration, and frequency precipitate their degradation. By integrating Earth Observation data into models, we can evaluate, on the one hand, the vegetation water stress and, on the other, its primary production. This allows us to assess the interaction of both processes, improving our knowledge about the vegetation's behavior in the face of drought.

 

In this work, we set up an open-source cloud framework to monitor water consumption and primary production interaction over this semiarid mosaic in the long term, to analyze system tipping points. This information can help reduce the uncertainty associated with the public administration and farmers’ decision-making processes. A surface energy balance model, previously validated in the area, was applied to estimate evapotranspiration (ET) from 2000-2020 (monthly, at a 1 km spatial resolution, using MODIS data and global atmospheric reanalysis dataset). The anomalies of evapotranspiration (ET) to reference ET were used as an indicator of drought for the period. The biomass production was estimated by applying an adaptation of the Monteith LUE (light use efficiency) model based on the relationship between plant growth and incident solar radiation. The parameterization of the model corresponded to semi-natural grasslands and crops, and it was applied at a daily scale with 250 m of spatial resolution. The model’s estimation presented an acceptable agreement over the area.

 

Close links between grassland/crop production and drought events were found and evaluated. 2016 was the worst year regarding the state of the vegetation, followed by 2015, 2003, and 2002, all coincident with drier events (as measured by ET/ETo anomalies). The different production patterns of each patch of vegetation were visible. Although crops were mainly rainfed (probably being irrigated if necessary) and followed the precipitation rates, they were less dependent on rain than grassland. Croplands had higher production peaks during February/March than natural grasslands, although trends were similar. Production rates were much higher than usual during 2004, 2009, and 2017. These vegetation blooms came after a drought where biomass production rates were minimal. A thorough analysis of these results can provide insights to better cope with future droughts.

Acknowledgment: This work has been carried out through the project "DroughT impACt on the vegeTation of South African semIarid mosaiC landscapes: Implications on grass-crop-lands primary production" funded by the European Space Agency in the framework of the "EO AFRICA R&D Facility".