Research And Application Of Air Temperature Spatial Estimation Based On Multi-source Information Fusion

The continuous spatial distribution data of near-surface temperature is of great value for scientific research on climate change,agricultural zoning,etc.The traditional spatial interpolation method can obtain the spatial distribution of regional temperature.However,due to the sparse and uneven distribution of ground meteorological stations,the temperature will be reduced.The spatial interpolation results have large errors,and only a limited range of temperature distribution can be obtained,and it is difficult to obtain near-surface temperature data with high spatial resolution.Satellite remote sensing will be covered by clouds,resulting in a large number of invalid pixels,which will affect the continuous spatial distribution of estimated temperature.In order to achieve high-precision near-surface temperature in continuous space,make up for the lack of weather station data,consider the impact of surface temperature and other factors,use remote sensing data and meteorological station measured data,establish a multiple linear regression model,and predict to get randomly generated The temperature value of the unknown location is then used as the soft data,and the measured data from the weather station is used as the hard data.The combination of the two is put into the Bayesian maximum entropy model to make a spatial prediction of the national average temperature in order to improve the wide range The air temperature is estimated in space and compared with the interpolation result of ordinary Kriging method.Taking Wuhan's heat island effect analysis as an example,the urban heat island effect of Wuhan is evaluated.The results show that(1)In terms of spatial distribution,the results of the Bayesian maximum entropy model combined with soft data can better reflect the difference in the details of temperature distribution and demonstrate the enhancement effect.(2)In terms of error evaluation,compared with the ordinary Kriging method,the MAE and RMSE of the Bayesian maximum entropy interpolation result combined with soft data are 0.95°C and 0.54°C smaller than that of the ordinary Kriging method,respectively.The prediction accuracy of the sparsely distributed western region is higher than that of the ordinary Kriging interpolation method.(3)Using the multiple linear regression model,the daily average temperature of Wuhan was estimated in space,and the third ring area of Wuhan was taken as the central area of the city,and 8 buffers of equal area were made outward.The results showed that,in general,Wuhan City shows a downward trend of temperature along the urban and rural areas,which may be related to the uneven distribution of waters and cultivated land in different buffer zones.The calculated Wuhan heat island intensity based on temperature is 0.95?.(4)The Bayesian maximum entropy method comprehensively uses multi-source data to realize the spatial estimation of temperature.The addition of soft data makes up for the lack of weather station data to a certain extent,and provides a new way to draw spatially continuous temperature distribution maps.