Improvements on the monthly precipitation spatial pattern characterization using a set of remote sensing products.

Abstract

Water availability is a limiting factor for many human activities and natural ecosystems processes. Monitoring of water resources, as well as the impacts of water scarcity on human and natural ecosystems, is key for defining adapted water management strategies. Currently, different European and Worldwide organisations are providing several climate services (CS) based on output datasets from weather forecast and climate projection models. To ensure the translation of these CS to actionable knowledge at a local scale, it has been required the tailoring and downscaling of data to fit the user requirements expressed by selected stakeholders representing different relevant sectors. This is one of the main goals of H2020 I-CISK project (https://icisk.eu/) which includes this study carried out at the Guadalquivir River Basin (RB) (small part of Guadiana RB), in the northern part of Andalusia, South of Spain. It is one of the seven established living labs (LL) in I-CISK. This LL is particularly vulnerable to drought impacts.The present work aims at evaluating the contribution of remote sensing data as an explanatory variable of the spatial pattern of precipitation, a key meteorological variable of water resources models. This characterization is a necessary preliminary step to understand the local relationships between climatic variables and others (topography, vegetation response, etc…) in order to subsequently apply known correlations to downscale  weather forecasting and climate projection models to the spatial resolution required by the user community.The method is based on the generation of multiple regressions with residual interpolation using weather stations' monthly precipitation data as the dependent variable and a set of independent variables at 250 m spatial resolution such as, squared distance to the Mediterranean Sea and to the Atlantic Ocean, elevation, cosine of aspect,  a set of remote sensing indexes (Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI)), synthetic versions of these indexes and corresponding anomalies.The NDVI used is generated by monthly aggregation of 16‐day MODIS composite products of MOD13Q1. NDWI has also been calculated from MOD13Q1 surface reflectance products. Synthetic NDVI and NDWI have been generated replacing the original pixel values by the neighbouring vegetation NDVI at the locations of gauge stations where land cover is categorized as impervious surface.  NDVI and NDWI anomalies are calculated based on the climatological monthly mean from the 2000-2021 MODIS data time series. Regressions include independent variables time lags of 0, +1, +2 and +3 months after with respect to the date of precipitation variable.Preliminary results of single year’s analysis show that including remote sensing data  to the analysis results in a better spatial characterization, obtaining higher correlations in the regressions, which are strongly dependent on seasonality. There is no clear pattern of which index (version and anomaly) is the best contributor and there is also no clear result for the response time lag between precipitation and the indices, although +2 months seems to be the most relevant. Future work will use a full time series analysis to obtain more information on these patterns.