High resolution rainfall distribution map can be derived from satellite observation, and rain gauges data can give us accurate rainfall measurements at a fixed location .Combined them together, the merging of satellite precipitation estimates and rain gauge measurement is to “calibrate” satellite precipitation estimation with ground observation, so that a more accurate rainfall field can be drawn. In this paper, we use a intellective objective analysis scheme which is brought forward by LU nai-meng (2004) to accomplish the merging or called “fusing”.
The satellite precipitation estimation data we used to fuse is from the National Satellite Meteorological Center (NSMC) which resolution is 0.1 degree. The rain gauge data is from the National Meteorological Information Center (NMIC). Based on the hourly and daily rain gauge precipitation observations of over 30000 rain gauges and the satellite precipitation estimation products, a real-time merging system to get the QPE product is established. Then we have done a series of evaluate and verify works to assess the accuracy.
There are mainly two kinds of original data sources. The satellite precipitation estimation data we used in this paper were obtained from the “FY-2E” weather satellite provided by the National Satellite Meteorological Center (NSMC) which horizontal resolution of 0.1 by 0.1 degree.
The evaluation precipitation data we used is the observed precipitation in the hydrological stations. The location of the hydrological stations was given in Fig2.b, which is obtained from the sharing data system between the China Meteorological Administration (CMA) and the Bureau of Hydrology, the Ministry of Water Resources of P.R.China (Bureau of Hydrology, MWR).
The satellite precipitation estimation product we used is based on the “split window channel technique” by NSMC. As well known, there are many kinds of typical algorithms of satellite precipitation estimation in the world , for example ,Barrett Method, Arkin Method, Life History Method, Stout Method, Griffith Method, Scofield Method, etc (Arnold Gruber,2000). But what we concerned technique is how to combine or “fusing” the satellite precipitation estimation product and the rain gauges data together. In this paper, we use a intellective objective analysis scheme which is brought forward by LU mai-meng (2004) to accomplish the fusing, which method is not only considered distance between the grid and rain gauges, but also consider the orientation of the rain gauges.The represents precipitation , superscibe to refers to observation,a refers to analysis and s refers to satellite estimate,the suffix is for the ith grid-point and is the raingauge location .the is weights and is the tatal number of raingauges.The is a distance factor , is a function of distance ,the is the orientation factor ,in which Where is a function of orientation. is the distance between station and grid point , is the average station distance is the angle from the connection line between station and grid to the connection line between grid and station .
In this study, the period from 20th May, 2016 to 10th Jul, 2016 in China were tested. First of all, the Satellite precipitation estimates data were downloaded, decoded, at the same time, the rain gauges precipitation data obtained, then the 2 steps was done:
①: Quality control. By a simple quality control method, we eliminated the irrationality rain gauges, then a relative immobile station list got.
②: Data supplement. Based on the step① and the daily and hourly rain gauges precipitation ,we supply the “zero” to the record in the station list in the step① .That means if the rain gauge doesn’t observed precipitation, we considered the precipitation in the rain gauge is “zero”.
In the fusing arithmetic, we selected 0.6° as the adjusting radius, which is the best adjusting radius after we have done a series of test from 0.1° to 2°( table ignoring).At last we have done a series of evaluation test by using the hydrological stations to examining the performance of the QPE products.
Evaluation was made with the threat score (TS), false alarm rate and missing alarm rate between the observations and the QPE products, and with the brier score to evaluate the performance. The precipitation was divided into four categories with the consistency of operational forecasts at the National Meteorological Center (NMC) of CMA as no rain, little rain, moderate rain and heavy rain with thresholds of 0.1mm (including), 9.9mm (including), 24.9mm (including) and 49.9mm (including), respectively. (Zhao et al 2010).
During the period of 20th May to 10th Sep. 2010, the threat score shows that precipitation accuracy was improved in all 8 regions and the whole china, the heavier the rainfall intensity, the more significant the improvement with the percentage between 20% and 60%. Meanwhile, bias score reveals decrease in all 9 regions; False alarm rate indicates obvious decrease for all 9 regions with a rate of 5% to 10%, for some region it is high up to 30%. Meanwhile, evident improvement was displayed by the missing alarm rate, especially for the shower and little rain, with a percentage of 30% for some regions.
As well known, there are obvious difference between satellite precipitation estimate and other remote technique as radar measurement, etc. But the results of satellite precipitation estimate can be widely used because the resolution of both spatial and temporal is good enough for the use of precipitation estimation .So, better results could be expected when the satellite precipitation estimate result fusing with rain gauges data and other data by multiple observed data.