| Precipitation is an important part of earth's water cycle and is closely related to the process of hydrology and meteorology.Study of precipitation is of great significance to the research of climate change,hydrological cycle and drought monitoring and so on.Precipitation products with high spatial and temporal resolution is very important for the accuracy of hydrological process simulation,hydrological cycle and drought monitoring.Traditional insitu rainfall stations are sparse and uneven in distribution,which is difficult to provide sufficient and accurate precipitation data.With the progress of remote sensing technology,the acquisition speed and data coverage of precipitation products have been increased,and precipitation products have been widely applied.The GPM satellite precipitation products process higher spatial-temporal resolution and larger coverage,which will have great application value in hydrological process simulation,drought monitoring and so on.However,there is still a problem that the spatial resolution of GPM precipitation products is a little rough in above application.Therefore,it is of great significance to study the spatial downscaling of GPM precipitation products and apply the downscaling products with high spatial resolution to the application of drought monitoring.The thesis takes the North China Plain as the research area,and aims at the low spatial resolution of the GPM rainfall data and the low monitoring accuracy of the traditional indices in the drought monitoring.The downscaling of the precipitation data based on the GPM precipitation and drought monitoring in North China Plain are carried out.The multi-source remote sensing data,including GPM and MODIS product,and in situ measured data of the research area are used.The paper appraised the applicability of GPM precipitation data quantitatively firstly.Then,a new downscaling model of GPM annual and monthly precipitation based on random forest algorithm is proposed by using precipitation covariate factors data,such as NDVI and LST.The annual and monthly GPM data was downscaled to the spatial resolution of 1 KM.At last,a comprehensive drought-monitoring index(CDI)is constructed by making full use of the 1 KM monthly precipitation data,which could provide more accurate support for the decision-making of the prediction and control of drought in the North China Plain.The main contents and conclusions of the thesis are as follows:(1)Validation of GPM precipitation product in North China Plain.The applicability of GPM product and the accuracy of single site are analyzed through the correlation analysis between GPM product and ground station precipitation data.In addition,the applicability of GPM product is compared with the TRMM product.The results show that the GPM product shows higher correlation and consistency with the in situ measured data on the annual,monthly and seasonal time scales.Especially in January,the GPM precipitation product is obviously better than that of TRMM data,indicating its improvement in observing slight precipitation.The single site accuracy of GPM product is also slightly better than that of TRMM data,of which 53.1%of the sites show higher correlation and consistency with the measured data.In conclusion,GPM precipitation product has high applicability and application value in North China Plain(2)Study on the downscaling model of GPM annual precipitation product.Effective covariate factor selection is carried out by analyzing the correlation between covariate factors and GPM precipitation.Then,a set of annual precipitation downscaling model based on random forest algorithm(RF)is constructed and the downscaling results is compared with the traditional single factor regression model(UR)and multiple factor regression model(MR).The results show that the R2 of three annual downscaling model(UR,MR and RF)are 0.7195,0.7221 and 0.8953 respectively,indicating that RF model could overcome the defects of the traditional UR and MR regression models.RF model is not sensitive to the multiple collinearity and is of less over fitting phenomenon,which could obtain the annual precipitation data of North China Plain more accurately.(3)Study on the downscaling model of GPM monthly precipitation product.Based on the downscaling result of the GPM annual precipitation,downscaling method for the monthly precipitation product is studied.Two models called annual-based fraction disaggregation model and monthly-based RF model are constructed and compared.The results show that the downscaling results of the two models are consistent with the in situ monthly rainfall data,but they are both of some overestimation phenomena.The monthly-based RF model shows a more serious overestimate,and its downscaling results presents a clear strip phenomenon in the spatial distribution.Also,the downscaling precision in the month with abundant precipitation(from May to August)was lower than that of the annual-based fraction disaggregation model.Overall,annual-based fraction disaggregation model performs better than monthly-based RF model in the downscaling of GPM monthly precipitation.(4)The construction of the comprehensive drought-monitoring index(CDI)based on the 1 KM monthly precipitation data obtained from the above downscaling model.The new index CDI is constructed by combining the traditational drought indices(VCI and TCI)and GPM monthly downscaling product.The monitoring precision is compared with the traditional drought indices(VCI,TCI and VHI).The results show that CDI had higher correlation with soil moisture data and higher drought monitoring accuracy that other indices.Drought was occurring more easily in late spring,early summer and late autumn;Spring drought is very severe,which is extremely harmful for the growth of the crops in North China Plain. |
PDF Full Text Request