Research on water level prediction for great lakes based on GAMLSS-GLO model

Abstract

The Great Lakes are the largest group of freshwater lakes in the world and are vital to many of the two countries' largest cities, supporting diverse uses and ecosystems. Although there are regulatory mechanisms of human control, natural phenomena such as rainfall and evaporation are still difficult to predict and control. Regional policies and environmental changes can also affect water levels, which in turn affect ecosystems and human communities. In this work, through the temporal analysis of short-term water levels, the researchers revealed the trend and pattern of water level change. Then, by comparing ARIMA model and PEARIMA model, the performance of different models in predicting water level is evaluated, and the accuracy of prediction is further improved and analyzed the data from 2000 to 2010. Further, we find the water level sequence autocorrelation coefficient and find the dependency in the confidence level of 95% is significant. Subsequently, we established the flow detection model of GAMLSS-GLO and random forest algorithm to detect the dynamic water flow of five lakes. From our extensive simulations, we can observe that our model can precisely predict the water flow changing in Great Lakes and provide reasonable management strategies for controlling the flows.