An adaptive weighted-average Kriging method applied to monitoring of freshwater ecosystems

Abstract

Context The prediction of freshwater quality is important for detecting pollution risks and assessing changes in freshwater ecosystems. As a high-precision interpolation method, Kriging was able to predict freshwater quality by using previously monitored data. However, how to select the key parameters, regression functions and correlation functions of Kriging method in the process of improving prediction accuracy is still a bottleneck. Aims This study aims to propose an adaptive weighted-average Kriging (AWAK) method to further enhance the accuracy of freshwater-quality predictions. Methods The AWAK method consists of four main steps. First, the key parameters influencing pollution indicators are selected by FPS method. Subsequently, six different Kriging candidate models are constructed by using regression and correlation functions with different characteristics. Then, an enhanced-likelihood function is used to determine the weights of the six Kriging candidate models. Finally, AWAK is built by weighted sum of these six Kriging models. Key results The AWAK outperformed traditional Kriging in predicting pH and dissolved oxygen, significantly reducing prediction errors. Conclusions By employing the AWAK method, this study successfully improved the accuracy of freshwater-quality predictions. Implications The introduction of the AWAK provides an effective approach in the field of freshwater ecology.