CLDAS陆面同化资料和ERA5-Land再分析资料气温产品在江西省的适用性评估

以江西省376个气象自动观测站的逐小时气温数据为基准，采用偏差、相关性和平均绝对误差等评价指标，对比分析2017—2022年CLDAS陆面同化和ERA5 Land再分析气温资料在江西省的适用性。结果表明： 1） ERA5 Land、CLDAS资料均能很好反映大部分站点的气温变化，CLDAS资料与观测资料的相关系数为0.99，相关系数区间分布较为集中；ERA5 Land资料与观测资料的相关系数为0.97，分布较为分散。2） 相较于观测站点多年平均气温，CLDAS资料较为接近，ERA5 Land资料则偏离较大。3） CLDAS资料的平均绝对误差明显低于ERA5 Land资料，二者均存在平原、盆地部分站点平均绝对误差较小而局部高海拔山区站点异常偏大的空间特征，以及秋季最大而冬季最小的季节特征。4） ERA5 Land资料偏差的日变化范围为-0.65—0.39 ℃，整体呈现单谷形分布；CLDAS资料偏差日变化范围为-0.05—0.05 ℃，波动幅度较小，没有明显的变化特征。5） 两种格点资料均能较好反映大部分站点的低温日数变化，但对于高温日数变化，ERA5 Land资料偏差较大，CLDAS资料偏差较小。

Applicability assessment of air temperature products of CLDAS land surface assimilation data and ERA5-Land reanalysis data in Jiangxi Province

Based on hourly temperature data derived from 376 surface automatic stations in Jiangxi Province, the ERA5 Land reanalysis temperature data and CLDAS land assimilation data from 2017 to 2022 year were analyzed by using mean deviation, correlation coefficient and mean absolute error (MAE) methods. Results showed that: 1) All grid datasets can well reflect the temperature changes in most of stations in Jiangxi Province. The correlation coefficient between CLDAS data and observation data was 0.99, with a relatively concentrated distribution. The correlation coefficient between ERA5 Land data and observation data was 0.97, with a more scattered distribution. 2) The annual average temperature of CLDAS data closely matched observation data, while ERA5 Land had greater deviation. 3) The MAE of the CLDAS temperature data was lower than that of ERA5 Land temperature data, all datasets exhibited significant spatial characteristics, with the MAE being lower in low terrain areas compared to high terrain areas. Furthermore, there were significant seasonal characteristics observed, with the largest MAE for temperature occurring in autumn, and the smallest in winter. 4) The daily variation of ERA5 Land temperature data exhibited a single valley pattern, with an average deviation of -0.65 to 0.39. In contrast, CLDAS temperature data was relatively stable with an average deviation of -0.05 to 0.05. 5) All of the datasets can well reflect spatial changes in low temperature days at the most stations. However, ERA5 Land data was found to have low reliability in reflecting changes in high temperature days, while CLDAS data showed high reliability in this regard.