Predicting plant community change using satellite remote sensing in the Greater Cape Floristic Region

• Earth observation enhances ecological monitoring by complementing field surveys. • Satellite data improves species abundance modelling beyond static environmental variables. • Joint species distribution models refine spatial prediction accuracy in semi-arid landscapes. • EO enables temporal tracking of plant community composition changes in the Greater Cape Floristic Region. • Findings support scalable biodiversity monitoring tools for conservation planning. Reliable landscape-scale monitoring is key to informing policy and guiding strategic intervention to halt and reverse global biodiversity loss. Time and resources constrain the scale of field surveys, but Earth observation (EO) satellites provide routine wall-to-wall coverage of the Earth. If changes in vegetation composition or abundance are captured in reflectance values, we can combine restricted field monitoring with EO data to improve our inferences at the landscape scale. We investigated whether EO data improved our capacity to predict the spatial distribution and temporal dynamics of vegetation in the Greater Cape Floristic Region (GCFR) of South Africa. Our analysis was based on the 211 most frequently observed plant species recorded in 1440 surveys. The value of EO was assessed based on the improvement to joint species distribution models (JSDMs) that were fitted with standard static environmental data. Topography and temperature were the most influential environmental drivers in both distribution and abundance models. The addition of EO resulted in a marginal increase in the explanatory power of distribution models (i.e., presence/absence) by 3%, while a more substantial enhancement was observed in species abundance models, with an increase of up to 30%. Nevertheless, the proportion of variance explained by EO was much greater, representing between 34% and 64% of the total. The inclusion of measurable EO variables replaced much of the residual variance that was otherwise explained by estimated spatial latent variables, allowing for more accurate predictions of composition across the landscape. We demonstrate that diverse field data combined with EO can substantially enhance the identification of spatial variation in abundance and temporal changes in the composition of highly diverse communities, such as those found in the GCFR.