Analysis of meteorological and hydrological evolution and establishment of runoff prediction model in Pailugou Watershed based on BO-LSTM

Abstract

[Objective] To reveal the characteristics of hydrological situation evolution in Pailugou watershed of Qilian Mountains, and to provide a basis and reference for future water resource management and optimal allocation in the watershed. [Methods] Based on the measured runoff and hydrological data of Qilian Mountain Field Observatory from 2000 to 2019, the effects of precipitation and temperature on runoff were investigated by using the linear trend method, Pettitt′s test, and wavelet analysis, et al., and a BO-LSTM runoff prediction model for the Pailudou Basin was established.[Results] 1)From 2000 to 2019, precipitation, air temperature and runoff in Pailugou Watershed showed a two-stage upward trend, and the cutoff point was in 2010, precipitation and runoff, the first stage of the upward trend are higher than the second stage, the slope is 10.74, 3.16 in turn; air temperature is the opposite, the second stage is higher than the first stage, the slope is 0.11. And precipitation, air temperature and runoff of the MK mutation test z-value are greater than 0.(2)Precipitation in the May-October months on the runoff changes of the contribution rate is larger; and air temperature in the December-April months on the runoff changes of the contribution rate is large.(3)The air temperature in Pailougou Basin mainly has two main cycles, 3 a and 14 a, of which the first main cycle is 14 a; runoff exists in three main cycles, 19 a, 9 a and 3 a, of which the first main cycle is 19 a; precipitation mainly exists in two main cycles, 4 a and 11 a, of which the first main cycle is 11 a.(4)The BO-LSTM runoff prediction model for Pailougou, with an accuracy of R² of 0.63 and a root-mean-square error of 14 047 m³, and the prediction accuracy of the model is greater in months with smaller runoff than in months with larger runoff. [Conclusion] Precipitation, air temperature and runoff in Pailugou Basin have been on an upward trend in the past 20 years. Runoff, precipitation and air temperature in Pailugou Basin have obvious cyclicity. Air temperature and precipitation are important factors affecting the runoff in Pailugou Basin. The runoff prediction model can be applied to Pailugou Basin. The above results provide scientific support for the study of water resource effects in the Qilian Mountains and the prediction of water resources in inland river basins.

Keywords

hydrology / water resources / runoff evolution / runoff prediction / Pailugou Watershed / neural network / LSTM model / Bayesain optimization algorithm