| How to obtain high precision and high accuracy rainfall data is an important research direction in the meteorology and hydrology and other related fields.At the same time,the rainfall information is also an important basis for natural disaster prevention and hydrological analysis and forecasting work.For many years,the weather station data are mostly used within the industry,it is not easy for ordinary researchers to obtain meteorological data with high spatial and temporal accuracy.Therefore,the technology of satellite detecting rainfall has become the focus of people's attention.Satellite technology began in the sixties of the last century,has only half a century of time,but in a short period of time to achieve a leap in development,the use of satellite to obtain high spatial and temporal resolution of rainfall data is entirely possible.There are currently more than ten kinds of satellites,can provide rainfall data,but the satellite data is still a big problem,such as its accuracy,can only be close to the value of ground rainfall,can not replace the accuracy of the rain gauge.The current rainfall product can reach 0.25 ° * 0.25 °,the space about 25 km,such precision is still too large for rainfall,and so how to obtain high-quality rainfall data has become a worthy study.This paper will study the satellite rainfall data in Sichuan Province,the results of the use of three kinds of satellite rainfall data are: CMORPH,TRMM and PERSIANN,these three data is currently using very popular rainfall data,the spatial resolution of 0.25 ° * 0.25.This paper will be divided into five parts.The first part is to analyze the information of rainfall in Sichuan Province for many years,so as to have a general understanding of the rainfall situation over the years in Sichuan Province.The second part is to analyze three kinds of satellites.The precision of the rainfall data in time and space will be divided into 3 hours,daily and monthly,so as to test the rainfall accuracy of the three data.The third is to combine the geographical factors including NDVI and latitude and longitude The third is to reduce the scale of the three kinds of data,hoping to get more accurate spatial rainfall information;the fourth is the combination of geoweighted regression method,in the addition and multiplicative fusion model,the satellite rainfall data and rainfall data,The fifth step is to test the fusion data obtained in the fourth step by using the ground rainfall data which is not involved in the fusion and verify whether the accuracy of the fusion data can be improved.The above five problems are studied,and the following results are obtained:(1)For the entire Sichuan province,rainfall from the selected 41 stations in the 30 years from 1981 to 2010 was about 900 more days in the summer than in the winter,and the fall was the same for all seasons In the smallest rainfall season.At the same time,rainfall in western Sichuan is much smaller than that in eastern Sichuan.In the 30 years,rainfall in Sichuan is mainly concentrated in two regions,one is in the east of Sichuan and the other is in Sichuan and Yunnan.In eastern Sichuan,the anomalies of rainfall were justified alternately before 1998,and after 1998,positive anomalies were more negative than negative anomalies.In the spring,rainfall in western Sichuan and eastern Sichuan increased over the 30-year period,with increases of 0.017 / 10 a and 1.3 / 10 a,respectively,while the rainfall in the other seasons was decreasing.(2)Since the spatial resolution of the three rainfall data is 0.25° * 0.25°,the rainfall gauge data in this spatial resolution is selected and averaged as the standard value of the satellite data.It is found that when the time scale is small,The results show that the correlation between rainfall and rain gauge is close to the result of CMORPH,especially for CMORPH,and the correlation coefficient between rainfall and rain gauge is 85%,87% and 87% in three years.respectively.The spatial distribution characteristics of CMORPH are very similar to those of the rain gauge.The spatial rainfall characteristics of TRMM and PERSIANN coincide with those of the rainfall meter in some areas of Sichuan.Of the scatter plot is also more discrete,correlation coefficient and deviation and root mean square error value is lower than CMORPH.At the same time,the variation trend of CMORPH daily rainfall is consistent with that of rain gauge,while PERSIANN is obviously different from that of rain gauge.(3)Before the validation of the three rainfall data,the response relationship between NDVI and NDVI was calculated first,and the response coefficient of NDVI to rainfall was the largest in 16 days.So,in the downscaling treatment,NDVI was 16 days Of the NDVI.After the downscaling,CMORPH was the closest to the rain gauge,and the correlation coefficient was also the highest among the three data,the deviation was below 2 for 3 years,the root mean square error was lower than that of the other two rainfall products The CMORPH downscaling data is similar to the rainfall data in the space,and there is no sudden change of the original data in the space,which shows that the downscaling data of CMORPH can reflect the rainfall information of the smaller spatial scale;TRMM,the downscaling trend of the downscaled TRMM is similar to that of the rain gauge,but the TRMM deviations are large in spring and winter,which is the same as that of the rainfall.The same is true for the PERSIANN data,and PERSIANN's rainfall trends are very different from those of the rain gauges,and do not reflect the true nature of daily rainfall variations.(4)On the basis of GWR,the three rainfall products were fused by the additive model and the multiplicative model,and a new fusion rainfall data with spatial resolution of 0.05 ° * 0.05 ° was obtained.The fusion data not only contained Satellite rainfall data also includes rainfall data and other geographic data,including latitude and longitude.The result of CMORPH is more accurate than that of multiplicative model in the whole Sichuan Province,and the result of CMORPH is similar to that of rainfall meter in some areas of Sichuan Province.The results of TRMM fusion are high in some time,but at some time,they are very different from the results of the gauges.In the paper,the fusion results of six PERSIANN were chosen.Except for the six months,the other fusion results of PERSIANN There is a big error.(5)Comparing the data with the precipitation data of 157 reference stations in Sichuan Province to verify the accuracy of the fusion data,the data of the base station can be obtained only from May to August of 2013,which is influenced by the difficulty of data acquisition.Are not involved in data fusion.It is found that the additive correlation model of CMORPH has the highest correlation coefficient,the least deviation and root mean square error,and the result of addition model is more accurate than that of multiplication model.The result of TRMM fusion is more accurate than the original data High,especially in the additive model,although the fusion results of TRMM in May 2013 and the base station there are differences,but in other months of the fusion results are still close to the base station;PERSIANN fusion results with the base station there is a big difference,The fusion result is not ideal. |
PDF Full Text Request