Assessing the Forecasting Accuracy of a Modified Grey Self-Memory Precipitation Model Considering Scale Effects

Abstract

Precipitation is an important parameter in water resource management, urban flood warning systems, and hydrological analyses. Precipitation forecasting can provide a decision-making basis for relevant organizations, such as those in the agricultural sector and water conservancy departments. In this paper, a modified grey self-memory model (MGSM) was constructed by combining a self-memory function and grey theory. To verify the precision of the model in cases in which measured data are not available in the forecasting stage, a self-test method based on the scale effect in the precipitation forecasting stage was proposed. Ultimately, the model was verified based on three precipitation scales—the annual scale, the crop growth period, and the monthly scale—in the crop growth period from 1961 to 2018 in the Songnen Plain area, Heilongjiang Province. The results showed that the MGSM yielded higher fitting accuracy than the original GM(1,1) and grey self-memory models. Furthermore, the precipitation in the study area was predicted with the MSGM at the three different scales above from 2019 to 2023. The accuracy of forecasting meets the relevant requirements, and the model can be used to forecast precipitation trends at different time scales in the future. The results provide a reference for formulating scientific and rational agricultural water use strategies and guiding agricultural production practices.