Research On Estimation Method Of Lake Storage Variable Based On Multi-source Remote Sensing Data

Lake water storage(LWS)change is a key factor to study global climate change response and watershed water cycle.Due to the limitation of natural conditions,it is difficult to popularize the traditional field measurement method,leading to the lack of storage data of most lakes in China.Remote sensing is free from these limitations and provides a feasible means for estimating LWS change on a large scale.However,the present remote sensing estimation methods have shortcomings in data quantity,data quality and model.To solve these problems,based on multi-source remote sensing images and satellite altimetry data,this thesis systematically studied the estimation method of LWS change from the aspects of method framework,data processing and model construction,and deeply analyzed the spatio-temporal variation of lake storage variables and their response to climate and human activities.The main research results include five aspects:1.Aiming at the problem of incomplete lake surface caused by cloud occlusion in the image.On the basis of summarizing the shortcomings of the traditional methods,this thesis proposes an interpolation algorithm for cloud occlusion region considering the boundary position and brightness variation.The designed water area interpolation operator can improve the accuracy of lake boundary location and lake area.The peak-shaped curve detection operator can eliminate the interference of image brightness change,extract more complete lake water region and improve the automation of water region extraction.The algorithm precision evaluation experiment gives the parameter setting of the algorithm,and compares the interpolation precision of different algorithms,which proves the effectiveness of the proposed algorithm.2.Due to the reflection interference of lake surface and surrounding ground objects,there are many anomalous elevation points of lake surface acquired by allocator satellite,especially the proportion of anomalous points of small and medium-sized lakes is higher,so it is difficult to extract effective elevation points of lake surface by traditional methods.Based on the analysis of previous research results,this thesis proposes an effective elevation point extraction algorithm of lake surface.Based on the stability and statistical law of lake surface,the algorithm selects the most elevation points within a certain elevation range as the effective elevation point set of the lake.The accuracy evaluation experiment gives the parameter setting of the algorithm,and compares the lake water level obtained by different algorithms.The results show that the proposed algorithm significantly reduces the uncertainty of lake water level.3.Hypsometric curve is a function representing the relationship between lake area and lake water level,which can be used to prediction of lake water level in the absence of satellite altimeter observations(extrapolating lake water level),but the current relationship curve construction process generally lacks the verification stage,and lacks the explanation for the rationality of predicting lake water level in long time series.This thesis presents a new way to select the optimal relationship curve by using the leave-one cross verification.This idea comes from the field of machine learning and deep learning,that is,the data participating in the curve fitting are divided into training set and verification set,and the curve with the best verification result is taken as the optimal curve,instead of using the training result(R~2 and RMSE)in traditional methods to evaluate the optimality of the curve.The test accuracy and cross validation results of the relationship curves with different sample sizes prove the rationality of this approach.4.The LWS change of four lakes in the study area were estimated using the above methods,and the rationality of the results was verified.First,the lake area and water level were extracted using Landsat 7/8 and Sentinel 2 images and Cryo Sat 2,ICESat 2 and Sentinel 3 altimetry data,respectively.Then,the optimal hypsometric curve was constructed and selected by least square regression and leave-one-out cross validation,and then the LWS change were estimated.Finally,the lake area and LWS change were verified and analyzed.Verification results show that the LWS change estimated in this thesis are reasonable.The spatial-temporal variation analysis shows that the lake area and water storage in the study area increased by 138.23±3.60 km~2 and 1.35±0.57×10~8 m~3 from 1987 to 2020,respectively.Before the 21st century,the lake area and water storage changed gently,and the overall change range was not large.But after the 21st century,the lake area and water storage increased sharply,especially since 2018,the LWS showed the strongest growth trend in 33 years.5.Based on the data of temperature,precipitation and cultivated land area,the potential factors affecting LWS change were discussed from the perspectives of climate and human activities.The results show that air temperature is one of the main factors affecting the long-term variation trend of lake water storage.In particular,the significant temperature difference before and after the 21st century can well explain the change of LWS during the same period.Precipitation is one of the main factors affecting the short-term fluctuation of LWS,and there is a significant positive correlation between the two,especially the heavy rainfall event in 2018,which is directly reflected in the sudden increase of LWS change in the same period.There was a significant positive correlation between cropland area and LWS change.Cropland area was an indirect reflection of irrigation water retreat,so the annual increase of irrigation water retreat might be the main factor affecting the increase of lake water storage.In addition,the annual increase in the amount of water returned from irrigation means the increase in consumption of fresh water resources and the deterioration of lake water quality.Especially,the study area is located in arid/semi-arid areas,which may aggravate the risk of water resources and bring great challenges to local water resources management.