Using cloud computing techniques to monitor long-term variations in ecohydrological dynamics of small seasonally-flooded wetlands in semi-arid South Africa

Abstract

• Semi-arid Southern African wetlands are lost due to unsustainable use and poor management. • Ramsar Convention on wetlands encourages development of frameworks for wetlands conservation. • The implementation of these frameworks is hindered by a dearth of high-resolution data. • Google Earth Engine (GEE) offers opportunities to address data challenges. • Findings underscores the relevance of GEE in evaluating wetland ecohydrological dynamics. Wetlands in drylands have high inter- and intra-annual ecohydrological variations that are driven to a great extent by climate variability and anthropogenic influences. The Ramsar Convention on Wetlands encourages the development of frameworks for national action and international cooperation for ensuring conservation and wise use of wetlands and their resources at local, national and regional scales. However, the implementation of these frameworks remains a challenge. This is mainly due to limited availability of high-resolution data and suitable big data processing techniques for assessing and monitoring wetland ecohydrological dynamics at large spatial scales, particularly in the sub-Saharan African region. The availability of cloud computing platforms such as Google Earth Engine (GEE) offers unique big data handling and processing opportunities to address some of these challenges. In this study, we applied the GEE cloud computing platform to monitor the long-term ecohydrological dynamics of a seasonally flooded part of the Nylsvley floodplain wetland complex in north-eastern South Africa over a 20-year period (2000–2020). The specific objectives of the study were 1) to evaluate wetland ecohydrological dynamics using the 20-year multi-date Landsat composite data coupled with the Random Forest machine learning algorithm, and 2) to establish the major drivers of wetland ecohydrological changes, using selected spectral indices (i.e. Normalised Difference Vegetation Index (NDVI), Normalised Difference Water Index (NDWI) and Normalised Difference Phenology Index (NDPI)) coupled with climate data. The ecohydrology of the wetland changed over time, with some classes increasing twice when compared to the previous measurement, while others decreasing significantly during the study period. Notably, the bare surface class increased at rates of 230% and 350% between 2006–2010 and 2016–2020, respectively. Moreover, the indices showed similar trends throughout the 20-year period, with NDWI having minimum values less than zero in all cases. This implied no surface inundation, although the presence of some wetland vegetation indicated seasonal to semi-permanent soil saturation conditions. A comparative analysis of climate data and remotely sensed indices showed that annual changes of precipitation and evapotranspiration were the main drivers of wetland ecohydrological variations. The findings of the study underscore the relevance of cloud computing artificial intelligence techniques, and particularly the GEE platform, in evaluating wetland ecohydrological dynamics for semi-arid southern African systems which are deteriorating due to the unsustainable use and poor management resulting from limited knowledge about their changes over time.