| High-temporal-resolution near-surface air temperature distribution data is of great value for scientific research such as climate change and agricultural zoning.Surface meteorological stations are sparsely distributed and it is difficult to obtain near-surface air temperature data with high spatial resolution.Because of the mutual influence from the underlying surface heterogeneity and other complex factors,the results of nearsurface air temperature spatial distribution simulation are not ideal.It is easy to observe the earth's surface in high spatial-temporal resolution with remote sensing technology,which makes it possible to accurately estimate and simulate long-time near-surface air temperature.In order to study the distribution of near-surface air temperature and accumulated temperature with high spatial-temporal resolution,taking the Huai River Basin as an example,MODIS image products and measured temperature data from 41 meteorological stations on the ground were used to select variables to establish a nearsurface air temperature estimation model.Daily land surface temperature data were reconstructed using methods such as Interpolation of the Mean Anomalies(IMA),Savitzky-Golay(S-G)filtering,and regression merging.Use the estimation model to obtain daily near-surface air temperature data.The five-day running mean method was used to determine the first day and the last day and calculate the spatial distribution data of 16 years of accumulated temperature.At the same time,the regional distribution differences and evolution characteristics of near-surface air temperature and accumulated temperature in Huai River Basin were analyzed.The main research results and conclusions are as follows:(1)The establishment of near-surface air temperature estimation model.Using the product data such as the daily land surface temperature of MODIS from 2003 to 2018 and the in-situ data,the optimal multiple linear regression model was obtained through correlation analysis and full subset regression analysis.The daily high,low and average daily temperature estimation models of the study area were established,R2 were 0.949,0.900,0.956,and RMSE were 1.997?,2.805?,1.866?.For a random forest regression model with the same variable,the corresponding R2 were 0.958,0.912 and 0.960,and RMSE were 0.809?,1.283? and 0.979 ?.Compared with the multiple linear regression model,the random forest regression model took longer time to run and the improvement of prediction effect was not significant.(2)Daily land surface temperature data reconstruction and near-surface air temperature spatial simulation.The reconstruction results showed that the land surface temperature reconstruction result at night was slightly better than the day,and it could better fill in the data loss caused by cloud cover.Use the reconstructed land surface temperature data and the obtained estimation model to simulate the spatial distribution of daily average temperature,maximum temperature and minimum temperature.The simulation results of daily average temperature and minimum temperature were good,and the maximum temperature was relatively poor.Compared with the in-situ data,the daily average temperature accuracy test results showed that R2 was around 0.91,and the RMSE and MAE were about 2.8 ? and 2.1 ?.Compared with the partial thin plate smoothing splines method,in terms of spatial distribution,the high and low values of daily average temperature and minimum temperature were in good agreement with the simulation results,and the maximum temperature deviation was slightly larger.The results of the partial thin plate smoothing splines interpolation method had large deviations in local details and were over-smooth in numerical values.While the temperature based on land surface temperature simulations retained good local details,but there were some abnormal points.In general,both methods have performed poorly at daily extreme values simulation.(3)Analysis of the characteristics of the spatiotemporal changes of the annual average temperature,seasonal average temperature and January average temperature in Huai River Basin.The results showed that the annual average daily temperature and the annual average minimum temperature presented a significant upward trend.The temperature change trend was in good agreement with the in-situ data,but the annual average maximum temperature increasing trend was not obvious.It was worth noting that the occurrence frequency of the extreme temperature in recent years has increased.The regional low temperature of Huai River Basin was mainly distributed in Yishusi River Basin and higher altitude area like Yimeng Mountain,Funiu Mountain,and Dabie Mountain.The regional high temperature was concentrated in the upper reaches of the Huai River far away from the coastline,local areas in Yangzhou and Taizhou,and south of the middle reaches of the Huai River.The annual average temperature had shown a general downward trend from the southwest to the northeast.The temperature in the Nansihu area and the south of the middle reaches of the Huai River had increased significantly.The overall temperature increase in Huai River Basin in spring was obvious,especially near the coast and northwest.The overall temperature increased in summer and autumn and the fluctuations in winter were large.The 0 °C isotherm in January was roughly along the line of Lianyungang Ganyu District-Nansi LakeShangqiu-Kaifeng.(4)Analysis of the characteristics of the spatiotemporal changes of accumulated temperature in Huai River Basin.It was found that the active accumulated temperature in Huai River Basin was not completely consistent with the annual average temperature.The accumulated temperature in most areas was in the range of 4500 ?·d-5000 ?·d.High-altitude areas,coastal areas,and high latitudes areas were the regions with low annual accumulated temperature.High-value were mainly distributed in the south of the Huai River and the southern area of Taizhou,Jiangsu.The 4500 ?·d active accumulated temperature contour had a tendency to move northward.And the maximum annual accumulated temperature showed a rising trend.Among them,the annual accumulated temperature in 2017 and 2018 had shown the maximum in recent years. |
PDF Full Text Request