[关键词]
[摘要]
以滦河流域为研究对象,对流域2018年GPM(global precipitation measurement mission)降水产品适用性进行评估;利用降水与海拔、经纬度和归一化植被指数(normalized difference vegetation index,NDVI)之间的相关性,构建基于PSO-BP(particle swarm optimization-back propagation)的GPM降水空间降尺度模型,得到2018年滦河流域空间分辨率1 km的降水数据。[JP]结果表明:原始GPM降水产品可以较准确地表达滦河流域降水,但总体存在降水量高估现象;以国家气象中心降水产品CGDPA(China gauge-based daily precipitation analysis)数据为基准,对空间降尺度前后的降水数据进行精度评估,发现在年、季和月不同时间尺度上,空间降尺度后的降水数据与CGDPA数据相关性更高,均方根误差更小,相对偏差值控制在±10%以内,[JP]这说明降尺度后降水数据空间分辨率和精度都更高。
[Key word]
[Abstract]
Accurate acquisition of precipitation data is very important for regional meteorological,hydrological,and water resources research.Precipitation data from meteorological stations are highly accurate,but the application of measured precipitation data is limited due to the uneven spatial distribution of stations.With the rapid development of satellite remote sensing technology,a large number of precipitation inversion products have developed.Precipitation products represented by TRMM are widely used in the fields of meteorology and water conservancy due to their high temporal and spatial resolution.Although remote sensing inversion precipitation products can effectively restore regional precipitation characteristics,there is still a problem of insufficient spatial resolution for refined water cycle simulation research. Taking the Luanhe River basin (LRB) as the research area,the applicability of the Global Precipitation Measurement Mission of 2018 was evaluated.Under the assumption of the unity of the relationship between data of different resolutions in the model simulation and based on the correlation between precipitation and impact factors that are elevation,latitude,longitude,and normalized difference vegetation index (NDVI),a product space downscaling model of GPM precipitation based on PSO-BP (particle swarm optimization-back propagation) was built.GPM precipitation products with a spatial resolution of 1 km in the LRB were obtained. In the evaluation of the applicability of GPM precipitation products,the correlation is only 0.59 at the annual scale.The correlation is the highest in spring and the lowest in summer at the seasonal scale.The root mean square error is the highest in summer and smaller in other seasons.The relative deviation values were greater than 0 on all time scales except spring.On the monthly scale,the correlation was greater than 0.6 in 9 months and reached more than 0.9 in April,May,and December.As June to September is the rainy season in the Luanhe River basin,the root mean square error was higher in July and August.The relative deviations were greater than 0 in all months except January,March,and April.Through analyzing original GPM and downscaling of GPM precipitation at different time scales,the correlation between the downscaling of GPM spatial data and the measured precipitation data increased to more than 0.85 at the annual and seasonal scales.The relative deviation values were close to 0 on all scales except winter.Compared with the precipitation measured by original GPM and downscaling GPM,the monthly correlation mean value increased by 0.25 and achieved above 0.8.The mean monthly root mean square error was reduced by 0.16 and the relative error was controlled within ±10%. The original GPM precipitation product can accurately express the precipitation in the LRB.However,there is a phenomenon of overestimation of precipitation and this is especially obvious in winter.At different time scales,the spatial downscaling of GPM precipitation based on the PSO-BP model has a higher correlation with the measured precipitation at the station.The root mean square error and relative deviation were smaller.This indicated that the GPM precipitation product obtained by the spatial downscaling method established could more accurately express the spatial distribution characteristics of precipitation in the LRB with higher accuracy.
[中图分类号]
[基金项目]