Blending long-term satellite-based precipitation data with gauge observations for drought monitoring: Considering effects of different gauge densities
Blending long-term satellite-based precipitation data with gauge observations for drought monitoring: Considering effects of different gauge densities
复制标题
将长期卫星降水数据与计量观测相结合以进行干旱监测：考虑不同计量密度的影响
DOI：
10.1016/j.jhydrol.2019.124007
复制
发表时间：
2019-10-01
影响因子：
6.4
通讯作者：
Wang, Peng
中科院分区：
文献类型：
--
作者：
Bai, Xiaoyan;Wu, Xiaoqing;Wang, Peng
将卫星降水估算 (SPE) 数据和现场测量观测数据相结合，可以生成有效的空间连续降水估算，且精度更高。本研究使用一种简单但有效的混合方法，即地理差异分析（GDA）方法，并使用人工神经网络-气候数据记录（PERSIANN-CDR）进行遥感信息降水估算作为案例研究，评估了与干旱监测现场测量观测相结合时长期SPE的改进情况。采用稀疏（50）、中（200）、稠密（727）三个不同密度气象站组的现场降水观测，评估仪表密度对SPE-仪表数据融合性能的影响。两个广泛使用的指数——标准化降水指数（SPI）和自校准帕尔默干旱严重程度指数（SC_PDSI）——被用作案例研究。除了稀疏的 50 个站点子集的情况外，SPE 仪表混合显示出原始 PERSIANN-CDR 数据的明显改进，无论是降水输入的准确性还是干旱监测的许多方面，例如干旱监测。再现干旱强度并揭示干旱的空间格局，其中SC_PDSI比SPI表现出更显着的改善。密集的 727 个站点集显示了混合数据的最大改进，但相应的仅站点插值也表现出与混合数据相当的性能，表明这些情况下 SPE 数据的利用率较低。只有中等密度 200 个站点集的混合结果显示出令人满意的干旱监测性能，并且相对于仅站点插值有显着改进。根据定量分析，中等密度（在我们的案例中，每10（6）km（2）约50-75个标尺）可能是SPE-标尺混合最经济的标尺密度，因为它对混合结果有令人满意的改善，可以充分利用SPE数据的优势，并且需要相对较少的标尺。我们的结果有助于了解SPE-计量混合如何改进基于SPE的干旱监测，并为数据有限条件下的干旱监测应用提供参考。后续的研究或应用还应仔细考虑规范密度的影响。
Blending satellite-based precipitation estimation (SPE) data and in-situ gauge observation data can generate effective spatially-continuous-precipitation estimates with improved accuracy. This study assessed the improvement of the long-term SPE when blending with in-situ gauge observations for drought monitoring, using a simple but effective blending method named the geographical difference analysis (GDA) method and with the Precipitation Estimation from Remote Sensed Information by using Artificial Neural Networks-Climate Data Records (PERSIANN-CDR) as case study. In-situ precipitation observations from three meteorological station sets with different densities-the sparse (50), medium (200), dense (727) station set-were adopted to evaluate the effect of gauge density on the performance of SPE-gauge data blending. Two widely-used indices-standardized precipitation index (SPI) and self-calibrating Palmer drought severity index (SC_PDSI)-were used as case studies. Except the case of sparse 50-station subset, the SPE-gauge blending shows apparent improvement to the raw PERSIANN-CDR data, for both the accuracy of precipitation input and many aspects of drought monitoring, e.g. reproducing drought magnitude and revealing spatial pattern of drought, in which SC_PDSI shows more significant improvement than SPI. The dense 727-station set shows the largest improvement in the blending data, but the corresponding station-only interpolations also exhibit comparable performance to the blending data, indicating lower utilization value of the SPE data for these cases. Only the blending results of the medium-density 200-station set shows satisfactory drought monitoring performance as well as significant improvements relative to the station-only interpolations. According to the quantitative analyses, the medium density (about 50-75 gauges per 10(6) km(2) in our cases) might be the most economic gauge density for SPE-gauge blending, as it has satisfactory improvement in blending results, can make fullest use of the advantages of SPE data and requires relatively fewer gauges. Our results can help to understand how the SPE-gauge blending could improve the SPE-based drought monitoring and serves as a reference for applying drought monitoring under the data-limited conditions. Subsequent studies or applications should also carefully consider the effect of gauge density.