Using historical normals to improve modern monthly climate normal surfaces for Madagascar

Abstract

High‐resolution, spatially continuous climate models are an essential first step in assessing fine‐scale impacts of climate and projected future climates. Observational weather station networks supporting such models are often inadequate and are spread unevenly throughout the globe, particularly in developing regions such as Madagascar, since network establishment and maintenance requires significant investment. Quality climate models are greatly needed for the assessment of conservation, agricultural productivity and socio‐economic studies of climate vulnerability.With the aim of generating improved modern climate normal surfaces for Madagascar, we have developed a method for augmenting the existing sparse climate normal network with spatial patterns of climate change derived from more densely distributed historical climate normals. Thin plate smoothing splines and kriging were applied to the spatial analysis of the climate difference patterns and their performances were compared.We found that the augmented data improved the spatial consistency of the climate interpolations when compared with the interpolations using only the existing observed data. Our analysis confirmed the robustness and accuracy of thin plate spline interpolation when applied to long‐term differences in climate normals. Further, the investigation indicated that long‐term differences in rainfall normals have broader spatial coherence and are less susceptible to local differences when compared to temperature normals. This suggests that long term change patterns in rainfall are better supported by sparse data networks, despite the usually acknowledged greater inherent spatial and temporal variability of rainfall.The augmented data capture distinct spatial patterns in temperature and rainfall that differ markedly from the patterns in interpolations of observed data alone. The derived change patterns are consistent with observed global trends. This technique might be used to improve understanding of the current climate in other data‐poor regions and the outputs used in spatial environmental and socio‐economic studies, such as species distribution mapping and climate change resilience mapping.