CLDAS和GLDAS土壤湿度资料在黄土高原的适用性评估

本研究基于2011～2013年3～9月CMA陆面数据同化系统(CLDAS)和全球陆面数据同化系统(GLDAS) Noah陆面模式提供的表层10 cm土壤湿度数据,以国家气象信息中心提供的站点土壤湿度观测资料为参考,通过对比分析两套模拟数据在黄土高原区域的时空差异,并分别计算其与观测资料的相关系数(Corr)、平均偏差(MBE)和均方根误差(RMSE)等统计特征值,就两套模拟数据在黄土高原地区的适用性进行综合对比和评估,旨在选出一种适用于研究黄土高原地区土壤湿度时空特征的大范围、长时间序列的替代资料。结果显示:(1)两套陆面模式资料均能较好模拟黄土高原地区土壤湿度的空间变化特征,主要呈现出从西北向东南和西南增加的趋势,其中CLDAS具有较高的空间分辨率,能够较好刻画研究区细部特征;(2)从站点角度的统计特征值来看,两套资料的Corr值普遍偏高,CLDAS有71%和63%的站点分别达到极显著和显著差异水平,而GLDAS的略低,分别为70%和62%;研究区内各套资料的MBE和RMSE分布均类似;正负偏差站点个数相差不大,分布区间主要为-0.05～0.05,其中CLDAS有26和32个站点分别处于-0.05～0和0～0.05之间,而GLDAS则为28和24; GLDAS的RMSE主要集中在0.05～0.07之间,而CLDAS绝大部分低于0.05;(3)从时间序列来看,GLDAS资料与实测最为接近,但在春季存在一定程度的偏干情况,偏干程度小于CLDAS;(4)从整个研究区土壤湿度的模拟状况来看,GLDAS的Corr、MBE和RMSE值分别为0.821、0.0126和0.0221,较CLDAS资料具有更小的平均偏差、均方根误差和更大的相关系数。总体来说,两套陆面模式资料在黄土高原区域土壤湿度模拟上均存在各自的优势,均可作为土壤湿度观测的替代资料,对于土壤湿度研究和业务应用都具有积极的意义。

Evaluation applicability of CLDAS and GLDAS soil moisture for the Loess Plateau

Field measurements of top 10 cm soil moisture in cropland were used to assess the applicability of the soil moisture derived from the Land Data Assimilation System to the Chinese Loess Plateau. Monthly soil moisture data from March to September during 2011 to 2013 derived from the CMA Land Data Assimilation System (CLDAS) and Global Land Data Assimilation System (GLDAS) were compared with field data, and the results were evaluated by using three indicators, correlation coefficients (Corr), mean bias errors (MBE) and root mean square (RMSE). The comparison was to select soil moisture data suitable for the condition on the Loess Plateau used in a large spatial and long-term temporal scale. The results showed that both the CLDAS and GLDAS soil moisture had the relatively high Corr, low MBE and RMSE.They could be used to reproduce the spatial variation of soil moisture over the Loess Plateau. However, the CLDAS soil moisture was found to be the best with relatively higher spatial resolution to better reflect the local minutiae than the GLDAS soil moisture. In terms of the statistical indicators, the Corr between field measurements and the two datasets was relatively high. Compared to the GLDAS soil moisture, nearly 71% and 63% of the gauging stations reached 0.01 significance and 0.05 significance level, while for the GLDAS soil moisture the proportion (70% and 62%) of gauging stations with 0.01 and 0.05 significance level was slightly lower; The spatial pattern of the MBE and RMSE for the two datasets were approximately the same across the Loess Plateau. MBE ranged from -0.05 to 0.05. Twenty six CLDAS gauging stations had values falling in -0.05~0 and 32 stations falling in 0~0.05. For GLDAS gauging stations these were 28 and 24 respectively. The RMSE of the GLDAS soil moisture was 0.05~0.07, while the majority of CLDAS was less than 0.05. But in terms of temporal variation of soil moisture and statistical indicators over the Loess Plateau, the GLDAS soil moisture better reflected dry and wet conditions. Overall both datasets had their own advantages over each other, and could serve as a good proxy for the soil moisture condition over the Loess Plateau.