不同再分析降水数据在洞庭湖流域的精度评估

收集了4种不同数据来源的再分析降水数据:欧洲中期天气预报中心大气再分析数据(ERA-intreim)、英国East Anglia大学气候研究中心数据(CRU TS3.21)、全球降水气候学项目数据(GPCP V2.2)、TRMM多卫星降水产品——TRMM 3B43(V7)和中国气象科学数据共享服务网提供的降水实测数据(CMD),用均值、偏差和偏差百分比及标准差等统计指标对再分析降水数据在洞庭湖流域的模拟精度进行评估。结果表明:(1)4种再分析数据年、季节降水量的年际变化与CMD总体相似,其中秋冬季节与CMD的一致性好于春夏季节。(2)ERA-interim数据对年均降水量的高估比例为20%以上,对季节平均降水量的高估比例为10.13%～33.65%,年、季降水量偏差百分比变化的年际变化较大;标准差的年内变化范围为12.40%～70.13%,数据集的离散程度最大。GPCP和CRU两种数据的年季降水量偏差在 -6.19%～5.71%之间变化,均值与CMD数据最为接近;CRU数据年季降水量偏差百分比的年际变化和标准差的年内变化均较GPCP的大。TRMM数据对年季降水量的高估比例为 3.67%～ 6.52%,降水量偏差百分比的年际变化稳定;标准差的年内变化范围为4.58%～14.03%,数据集的离散程度最小。(3) ERA-interim、CRU、GPCP和TRMM数据年降水量的偏差范围分别为 1 000.21～-649.51、 333.89～-643.82、292.55～-686.85、256.20～-561.27 mm。再分析降水数据对降水的低估区域均出现在以南岳站为中心的衡阳盆地,而高估区域分布则有明显差异:ERA-interim的高估极值出现在地势较高的西部,CRU和GPCP的高估极值出现在南部山地,TRMM无明显高估区域。ERA-interim数据离散程度的空间分异明显,CRU和GPCP居中,TRMM分异较不明显。所有再分析数据在西部和南部山地区域的离散程度最大。

EVALUTION OF PRECIPITATION FROM ERA-INTERIM,CRU, GPCP AND TRMM REANALYSIS DATA IN THE DONGTING LAKE BASIN

To find a proper reanalysis dataset which could best agree with the precipitation characteristics of the Dongting lake basin (107°13'-114°18' E、24°35'-30°27' N), reanalysis precipitation data from ERA-intreim, CRU-TS3.21, GPCP V2.2 and TRMM 3B43(V7) were collected to be compared with the actual observation precipitation from CMD to assess their fitting and dispersion degree, by the analysis of mean, precipitation difference (PD), the percentage of precipitation differences (PPD) and standard deviation of PD (SD-PD) and PPD (SD-PPD). All the four reanalysis precipitation datasets were interpolated to the same 27 meteorological stations as CMD provides. The results indicated that: 1) annual and seasonal variations of the four reanalysis precipitation datasets were consistent with that of CMD, while the reanalysis precipitation data in autumn and winter fitted more closely with CMD than other seasons did;2) during 2000-2012, the average annual precipitation in Dongting lake basin of ERA-Interim, CRU, GPCP, TRMM, and CMD, are 1 648.71, 1 322.87, 1 317.76, 1 403.13 and 1 332.69 mm, respectively. Correspondingly, PPDs between four reanalysis precipitation datasets and the CMD data were 23.71%, -0.74%, -1.12% and 5.29%. In addition, spatial distribution of PD between them ranged from 1 000.21 to 649.51 mm, 333.89 to 643.82 mm, 292.55 to -686.85 mm, 256.20 to -561.27 mm, respectively, while their SD-PDs ranged from 445.89 to to 97.35 mm, 365.53 to 48.66 mm, 251.42 to 43.12 mm and 204.11 to 32.76 mm;3) of the four reanalysis precipitation datasets, ERA-interim exhibited the greatest overestimation on value and on time scale during the two periods, and its annual variation of PPD and dispersion degree were the largest. The average precipitation of CRU was more similar to CMD than that of GPCP,while the interannual variation of PPD as well as dispersion degree of GPCP were smaller than those of CRU. TRMM overestimated the precipitation slightly on time scale, and possessed stable interannual variation and minimum dispersion degree;4) precipitation in Hengyang basin, with Nanyue station as the center, were underestimated by all reanalysis datasets. The spatial distribution of overestimated value had significant difference. Overestimated extremum for ERA-interim occurred in western mountain area, however, for CRU and GPCP, in southern mountain area, and the overestimated area was not obvious for TRMM. There was distinct difference in spatial distribution of dispersion degree for ERA-interim data, followed by CRU, then GPCP, and there was no significant difference for TRMM. The maximum dispersion degree of all reanalysis data existed in western and southern mountain area of the Dongting lake basin. According to the analysis above, although TRMM data overestimated the mean value slightly, it was found that TRMM data presented the best consistency with CMD data than GPCP and CRU data did, and ERA-Interim data was the worst.