Construction And Application Of Inversion Model Of Lake Surface Water Temperature

The sixth assessment report of the Intergovernmental Panel on Climate Change has pointed out that the global average surface temperature has risen by 0.99? in the past 20 years.Lakes are also characterized as an important part of aquatic ecology,and their Lake Surface Water Temperature(LSWT)is on the rise as a whole.As one of the most important indicators of lake ecology,lake surface water temperature not only affects the biophysical and chemical processes in the lake,but also affects the watershed where the lake is located and its surrounding environment.Human activities are closely related to changes in lakes.Therefore,the study of lake surface water temperature can better realize lake governance,human and ecological symbiosis,and ecological balance.However,the late start of in-situ observation of lake surface water temperature,sparse sites,scarce measurement data,and insufficient research on the spatio-temporal scales,the relative satellite remote sensing image data is sufficient but cannot directly obtain the lake surface water temperature data,which limits the change of lake surface water temperature in the past historical development and the future development trend of the lake surface,and it is urgent to build a model for lake surface water temperature based on satellite remote sensing image data to realize the quantification of lake surface water temperature data,based on which to achieve inversion and prediction of lake surface water temperature.In this study,11 typical plateau lakes in the Yunnan-Guizhou-Sichuan Plateau were selected as the research objects.Based on the Landset and MOD11A2 remote sensing image data from 2001 to 2018,this study constructs a model of lake surface water temperature based on satellite remote sensing image data and realizes the inversion and prediction of surface water temperature of each lake.(1)Combine the Air2 water model and the long-short term memory(LSTM)to design an LSWT prediction model to achieve LSWT inversion and prediction for the study area.(2)Root mean square error(RMSE),mean absolute error(MAE),standard deviation(SD),regression analysis and correlation analysis were used to analyze the model performance.(3)Construct LSWT data of 11 lakes to 2025 based on existing datasets and prediction models,and complete trend analysis and cycle analysis of LSWT datasets from 2021 to 2025.The research results show that the fitting degree of Air2 water model and LSTM model to estimated data and observation data is more than 0.8,while the Air2 water model(RMSE: 1.04°C,0.76°C and 0.28°C)on monthly,seasonal and annual scales ?MAE: 0.89?,0.66? and 0.25? SD: 0.93?,0.61? and 0.25?)and LSTM(RMSE:1.23?,0.86? and 0.40? MAE: 1.03?,0.75? and 0.36? SD: 0.79?,0.54 °C and 0.28 °C),the three average error values are all small,indicating that the two models have good performance in inversion and prediction of LSWT,so the progress of the LSWT prediction model constructed based on the two models is reliable.The prediction results show that 6 lakes show significant warming(up to +0.2°C/year),and the remaining 5 lakes show no significant change.Classify based on the performance of the two models at different temporal and spatial resolutions,and construct monthly,quarterly and annual LSWT data sets from 2021 to 2025 based on the prediction results.It is found that the size classification of LSWT data on different spaces is consistent with the classification of their altitude.Through the further exploration and research of physical process model and machine learning model in LSWT inversion,this study provides new solutions for LSWT prediction and inversion,and provides strong basic theoretical and practical support for subsequent LSWT research,which has certain practical significance and research value.