Palmer, AR; Ezenne, GI; Choruma, DJ; Gwate, O; Mantel, SK; Tanner, JL (2020). A comparison of three models used to determine water fluxes over the Albany Thicket, Eastern Cape, South Africa. AGRICULTURAL AND FOREST METEOROLOGY, 288, 107984.

The Albany Thicket (AT) biome contains outstanding global biodiversity as well as the potential to achieve carbon credits associated with water-efficient Crasslucean acid metabolism (CAM). Understanding the water fluxes in the AT is crucial to determining carbon (C) sequestration rates and water-use efficiency. Despite large variation in water fluxes across the AT, only a few studies have been conducted in this region with their results validated against short periods of observed data. This study aims to evaluate three models of water fluxes over AT against data from an eddy covariance (EC) system active from October 2015 to May 2018. ET was modelled using the BioGeoChemistry Management (BGC-MAN) model, a biophysical model (Penman-Monteith-Leuning (PML)) and a remotely-sensed product (MOD16), and their results compared with that from the EC system. More than three decades of rainfall data from Climate Hazards Group InfraRed Precipitation with Station Data (CHIRPS) was used to assess some rainfall characteristics of the region. The mean annual rainfall is 404 mm and mean monthly rainfall ranges from 16.0-50.7 mm, with minima likely to occur in winter period (between May and July) and monthly maxima in the summer period (between October and March). Among the three hydrological years in this study, total ET for 2016-2017 exceeded rainfall received by about 7% which shows that AT is likely to be supported by groundwater at some point but this requires further investigations. Generally, the three models applied in this study performed reasonably well when compared with the measured ET. The cumulative ET from BGC-MAN was slightly higher than that from EC by 16% and 8% in 2015-2016 and 2017-2018 hydrological years respectively while PML was slightly lower by 3% and 17% in 2016-2017 and 2017-2018; additionally, MODIS was slightly lower by 14% and 7% in 2016-2017 and 2017-2018, respectively. However, the correlation between the ET from EC and simulated ET from the three models was significant at p < 0.01.