Hydro-geomorphic dynamics in the Makgadikgadi Okavango Zambezi Basin, Northern Botswana

Abstract

The Makgadikgadi-Okavango-Zambezi (MOZ) basin of Northern Botswana is examined through a series of connected studies at different spatio-temporal scales. With an aim of using different techniques to look at changes and drivers of change in hydrology where the MOZ terminates (Mababe, Ngami, Makgadikgadi), this work looks at dynamics from three perspectives in three chapters: 1) a historical approach (chapter) that examines changes in hydrology and causal factors over the last 200 years; 2) an examination of the soils and their origins (parent materials) in the terminal basins; 3) a case study using Lake Ngami to monitor seasonal changes to surface water over the last 20 years. For the historical chapter, reports and surveys including photographs were used and digitized to complete a catalog of changes to hydrology. Soil pits were excavated at 31 locations and to further investigate provenance elemental analyses, radiocarbon dating, and carbon isotope analyses were conducted. For analysis of change in the areal extent of Lake Ngami Google Earth Engine was used to review and calculate indices from complete Landsat and Sentinel archives. Results from the historical approach provide new photographic evidence combined with longitudinal stream profiles that shed light on the floods of 1925, previously thought to be the largest in this area. Soils that were radiocarbon dated confirm with other studies that 3000 ybp Lake Ngami was much larger and some massive shift in the system caused the regular seasonal inflows to the lake to cease. Similar lacustrine deposits were not found in excavations from either the Makgadikgadi or Mababe basins at topographic lows where deposits would have been expected if these were all filled simultaneously through a connected system. Water indices created for Lake Ngami over the last 20 years were comprised of over 100 scenes from 3 sensor platforms. Key findings include a direct correlation between water indices derived from atmospherically-corrected Landsat-8 and those derived from Sentinel-2 that were not atmospherically corrected. This finding allows for more complete time-series to be constructed in the future without the need for further processing as previously indicated in the literature.