Study On Scale Effect Of Precipitation Based On TRMM Data In The Tianshan Mountains

Precipitation is one of the important links in the global material and energy cycle,and is also an important evaluation index for regional ecological and hydrological changes.Its data is used as a key parameter in relevant studies such as climate change,water resources evaluation,and hydrological models.Precision.With the advancement of remote sensing technology,TRMM precipitation data is widely used in regional hydrological research,but the resolution of this data is only 0.25°×0.25°,which is not enough to accurately describe the distribution of precipitation in complex terrain areas.Spatial downscaling technology can provide technical support for improving the spatial resolution and accuracy of TRMM data,but mountainous areas with complex geomorphic patterns and meteorological conditions have spatial differences in the scale effect of precipitation data.For this reason,the scientific construction of spatial downscaling models and the reasonable selection of data scales are particularly important.This is not only a useful exploration of downscaling related technologies,but also a useful supplement to the meteorological station scarce mountain precipitation data.This study takes the Tianshan mountainous area as a research sample area,supported by geographic information technology,and based on the analysis of TRMM precipitation data suitability,using partial least squares(PLS),random forest(RF),and random forest-partial least squares(RF-PLS)three methods to construct a spatial downscaling model,and obtain precipitation distribution data with four spatial resolutions of 90m×90m,250m×250m,500m×500m,and 1km×1km By comparing and analyzing the scale effect,the best spatial scale of TRMM data is determined.Finally,the characteristics of the spatial and temporal distribution of precipitation in the study area and the reasons for its formation are explored in combination with the topography and other features.The main research conclusions are as follows:(1)Through the comparative analysis of TRMM inversion and site-measured precipitation data,it is found that the overall change trend of the two is relatively consistent,the determination coefficient(R~2)is 0.88,the root mean square error(RMSE)is 0.37,and there is a significant linear correlation,indicating TRMM The precipitation data has good applicability in the Tianshan Mountains.(2)The accuracy of the three spatial downscaling models was verified,and it was found that the accuracy of the simulated data was improved compared with the original TRMM precipitation data.Although the simulation accuracy of the three models is different in different months and elevation zones,the overall accuracy of the RF-PLS model is the highest,R~2reaches 0.9,and the RMSE is 0.43.To this end,this paper selects the RF-PLS downscaling method to carry out the simulation study of the multiscale effect of precipitation in the Tianshan Mountains.(3)Based on the RF-PLS spatial downscaling model,the spatial resolution of the original TRMM precipitation data is increased to four different scales.The fit between each downscaling result and the original data is analyzed by comparison,and the year,season,and month are comprehensively analyzed.The elevation,the slope and the aspect direction of the simulation results of different time and space scales,the best scale of precipitation simulation in the study area is 500m×500m,its R~2 is 0.94,the RMSE value is 0.39,and the simulation scale of other scales is in order 90m×90m,250m×250m,1km×1km.(4)Based on the precipitation data of the best scale(500m×500m),the spatial-temporal distribution characteristics and the causes of precipitation in the study area were analyzed.The order of precipitation in different geomorphic zones is as follows:extremely high mountain>middle high mountain>high mountain>middle mountain>middle low mountain>plain>low mountain>basin;the ratio of precipitation in the three regions of basin,plain and mountain is about 1:2:4;The maximum rainfall values appear in the middle and high mountainous areas and the extremely high mountainous areas,respectively.There is very little precipitation in the Tuha Basin in the southeast of the study area,and the average monthly precipitation is less than10mm for many years.The Ili Valley is within 400mm and other precipitation lines.The precipitation in this area is abundant.