A spatial interpolation based on neighbor cluster adaptive model with spatial color block clustering algorithm

Abstract

Accurate soil nutrient data are crucial for precise fertilizer recommendations in intelligent agriculture. However, the process of soil testing, which includes collecting samples, determining available nutrients and interpreting results, is expensive. To address this challenge, spatial interpolation methods are commonly used to predict soil fertility. Yet, existing techniques like IDW (Inverse Distance Weighting) and OK (Ordinary Kriging) face limitations, making it difficult to achieve highly accurate estimates. Therefore, this paper introduces NCAMS (Neighbor Cluster Adaptive Model with Spatial Color Block), a novel interpolation approach that automatically identifies nearby points crucial for estimating soil nutrient values at a given location. In our approach, we not only consider spatial correlation but also incorporate the soil variables of sampled points. Delaunay triangulation and hash functions further divide data points into distinct clusters, with our model automatically selecting specific clusters. Moreover, our interpolation method integrates IDW and OK without requiring extensive training on real-world data. Extensive experiments on four real-world datasets, conducted through cross-validation, demonstrate the superior performance of our approach compared to eight state-of-the-art methods.