Geostatistical Analysis for Rain Gauge Network Design Based on Remotely Sensed Agroclimatic Zoning Towards Sustainable Agriculture

The required amount of water for crop production differs from region to region, mainly due to the decreasing “effectiveness” of rainfall in maintaining plant growth due to increasing evaporation. Effective rainfall is related to the available moisture in the plant’s root zone, allowing the plant to germinate, emerge, and maintain its growth. There is, thus, a need to design agrometeorological and rain gauge networks for monitoring suitable crops, leading to sustainable agriculture. However, the minimum density needed for a rain-gauge network depends on the variability in space and time of the intended precipitation fields, and a variety of factors, including topographical features, accessibility, and ease of maintenance, influence the choice of rain-gauge locations. The number of rain gauges, rather than their location, is what optimizes a rain gauge network for various time scales, such as daily or monthly data. To map mean daily or mean monthly precipitation over watersheds or medium-sized agroclimatic zones and monitor crops for sustainable agriculture, this study focuses on the number of grid cells leading to the design of a rain gauge network, specifically a regular rain gauge grid scenario, considering the areal extent of land suitable for agriculture based on the agroclimatic zoning methodology. The universal kriging approach, an interpolation method based on geostatistical analysis, is used to calculate the required number of grid cells for the network’s density. Geostatistical analysis focuses on estimating and selecting the semi-variogram (SV) to determine the range or the decorrelation distance between rain gauges for precipitation estimation, superimposed on a grid mesh that encompasses the area’s agricultural land, such as Thessaly in Greece, as delineated by the agroclimatic zoning methodology. The daily and monthly mean climatological precipitation database is used, based on the CHIRPS-v3 dataset. The results indicate that the best fit for both daily and monthly mean precipitation is provided by the Spherical SV model, and the corresponding maximum range, which also represents the decorrelation distance, is calculated at 28.162 km for both mean daily precipitation and mean monthly precipitation, respectively. In summary, the size of the selected grid cells, based on the computed range of the selected SV, is 28 × 28 km2 for the region of Thessaly, where one rain gauge is expected to be installed in each grid, leading to a total number of 18 grid cells.