A multicriteria decision analysis framework for generating baseline virtual weather stations in Zimbabwe

Tinetariro Chikati , Tambudzai Theresa Ngatiane

In this study, we generate baseline Virtual Weather Stations (VWS) in locations with minimal atmospheric noise by avoiding two major surface factors known to cause noise in remotely sensed weather observations. The study also proposes an improvement to the spatial coverage of Zimbabwe's weather observation network through the identification of environmentally suitable locations for virtual station deployment, where physical infrastructure could also be placed. A geospatial multi-criteria decision analysis approach was applied, using a weighted combination of Terrain Ruggedness Index (TRI) and proximity to large-scale water bodies, constrained by distance from existing physical weather stations, to determine optimal locations. Sixty-three Virtual Weather Station locations were generated to complement the existing network of 45 active physical weather stations. To maximise coverage and reduce duplication of observations, the virtual stations were strategically located beyond the influence zones of existing stations. The average nearest-neighbour distance of 59.68 km indicated a well-balanced and spatially efficient network. Network representativeness was further assessed across Zimbabwe's Agro-Ecological Zones (AEZs). The results showed that Regions IV and Va received the highest number of virtual stations. Conversely, the existing physical weather station network was concentrated within smaller AEZs, leaving larger regions relatively underrepresented. Environmental validation demonstrated that the optimised network occupied significantly less rugged terrain than a randomly generated control network. The VWS locations recorded a mean Terrain Ruggedness Index of 10.23 m and a median of 9.33 m, compared to 53.79 m and 29.82 m for the random network. A Mann–Whitney U test confirmed that these differences were statistically significant (U = 425.00, p < 0.001). Blind spot analysis further revealed that integrating the virtual network with existing stations substantially reduced observational gaps across Zimbabwe, demonstrating potential to improve weather monitoring in regions with sparse physical monitoring infrastructure.