基于CMFD 2.0和SPEI的江苏省1951—2024年气象干旱事件时空演变特征

Spatiotemporal Evolution of Meteorological Drought Events in Jiangsu Province from 1951 to 2024 Based on CMFD 2.0 and SPEI

摘要:

干旱是江苏省最主要的气象灾害之一,准确评估区域干旱时空演变特征对农业水资源管理和防灾减灾具有重要意义。基于第二代中国区域高时空分辨率气象驱动数据集（CMFD 2.0）、江苏省内13个常规气象站以及3个太阳辐射观测站逐日气象数据,采用FAO56 Penman-Monteith（P-M）模型和Priestley-Taylor（P-T）模型估算参考作物蒸散量（ET0）,评估CMFD 2.0产品在江苏省的适用性,分析CMFD 2.0产品结合P-T模型估算ET0精度,并评估CMFD 2.0降水产品误差在月尺度标准化降水蒸散指数（SPEI）计算以及干旱事件识别中的传递,最后结合改进的Mann-Kendall方法从年际变化、季节分异和空间格局3个维度分析江苏省气象干旱频率、历时和强度的时空演变特征。结果表明,CMFD 2.0日尺度最高气温、最低气温、太阳辐射及月尺度降水产品具有较高精度,决定系数（R2）为0.987、0.950、0.871及0.960,均方根误差（RMSE）为1.30℃、2.17℃、2.95 MJ/(m2·d)和18.18 mm,但日尺度降水产品精度相对较差（R2=0.648,RMSE为6.16 mm）。基于CMFD 2.0驱动P-T模型估算的ET0与基于站点实测气象数据驱动P-M模型估算的ET0具有良好一致性（R2=0.88,RMSE为0.81 mm/d）。基于CMFD计算的SPEIcmfd与站点SPEIsta结果高度一致（R2=0.942,RMSE为0.24）,干旱等级识别总体分类精度（OA）为87.72%,Kappa为0.760,误差主要集中于相邻等级,整体不影响区域干旱监测可靠性。江苏省1951—2024年平均SPEI无显著变化趋势,但季节差异明显,春季SPEI呈显著下降趋势,平均变化率为-0.083/（10a）,其余季节平均SPEI变化趋势均不显著。各等级干旱发生频率由大到小为轻旱（15.6%）、中旱（9.5%）、重旱（3.8%）、特旱（2.1%）,2月发生干旱频率最高（36.0%）,10月发生干旱频率最低（30.7%）。干旱次数和累计历时呈北高南低的空间格局,干旱强度表现出更复杂的局地差异,苏北地区为干旱敏感区,而苏南地区干旱更多表现为低频、高烈度事件。在季节上,全省春季干旱发生频率空间差异最大（2.7%~60.2%）,夏季空间差异最小（23.9%~43.2%）。春、秋和冬季干旱频率呈单一纬度梯度,在研究时段内以上升趋势为主,其中春季上升趋势达显著水平的面积占比为52.7%,而夏季干旱频率以下降趋势为主。干旱历时和强度均呈北部上升以及南部下降的变化特征,苏北地区干旱风险较高,需加强监测预警。

Abstract:

Drought is one of the most severe agro-meteorological hazards in Jiangsu Province, China. Accurate characterization of its spatiotemporal evolution is essential for agricultural water resources management and drought risk mitigation. Using the second-generation China Meteorological Forcing Dataset (CMFD 2.0) and daily meteorological observations from 13 normal weather stations and three solar radiation stations in Jiangsu Province, reference evapotranspiration (ET0) was estimated by using the FAO-56 Penman-Monteith (P-M) and Priestley-Taylor (P-T) models. The applicability of CMFD 2.0 in Jiangsu Province was evaluated, and the accuracy of ET0 estimated by coupling CMFD 2.0 with the P-T model was assessed. Monthly standardized precipitation evapotranspiration index (SPEI) values were then calculated by using the estimated ET0 and CMFD 2.0 precipitation data. Combined with the modified Mann-Kendall test, the spatiotemporal evolution of meteorological drought frequency, duration, and intensity in Jiangsu Province was analyzed from the perspectives of interannual variability, seasonal differences, and spatial patterns. The results showed that the CMFD 2.0 air temperature and solar radiation datasets exhibited high accuracy, with coefficients of determination (R2) of 0.987, 0.950, 0.871, and 0.960 for daily maximum temperature, daily minimum temperature, daily solar radiation, and monthly precipitation, respectively. The corresponding root mean square errors (RMSEs) were 1.30℃, 2.17℃, 2.95 MJ/(m2·d), and 18.18 mm. In contrast, the daily precipitation dataset showed relatively lower accuracy (R2=0.648, RMSE=6.16 mm). The ET0 estimated by using the CMFD 2.0-driven P-T model agreed well with that estimated using the station-observation-driven P-M model (R2=0.88, RMSE=0.81 mm/d). The SPEI derived from CMFD agreed well with that derived from station observations (R2=0.942, RMSE=0.24). The overall accuracy (OA) and Kappa coefficient for drought category classification were 87.72% and 0.760, respectively. Most classification errors occurred between adjacent drought categories, suggesting that they did not substantially affect the reliability of regional drought monitoring. The annual mean SPEI showed no significant trend during 1951—2024, whereas significant seasonal differences were observed. Spring SPEI exhibited a significant decreasing trend at a rate of -0.083 decade-1, while no significant trends were detected in the other seasons. The occurrence frequencies of mild, moderate, severe, and extreme droughts were 15.6%, 9.5%, 3.8%, and 2.1%, respectively. Drought occurred most frequently in February (36.0%) and least frequently in October (30.7%). Drought occurrence and cumulative duration exhibited a clear spatial pattern of being higher in northern Jiangsu and lower in southern Jiangsu, whereas drought intensity showed more complex local variations. Northern Jiangsu was identified as the most drought-prone region, while droughts in southern Jiangsu were characterized by lower frequency but higher severity. Among the four seasons, spring showed the largest spatial variation in drought frequency (2.7%~60.2%), whereas summer exhibited the smallest variation (23.9%~43.2%). Drought frequency in spring, autumn, and winter generally increased during the study period, with 52.7% of the study area showing a significant increasing trend in spring, while drought frequency in summer was dominated by a decreasing trend. Both drought duration and drought intensity increased in northern Jiangsu but decreased in southern Jiangsu, indicating that northern Jiangsu faced a higher drought risk and required strengthened drought monitoring and early warning.