Yu Manzhu, Yang Chaowei
NSF Spatiotemporal Innovation Center and Department of Geography and Geoinformation Science, George Mason University, Fairfax, VA, United States of America.
PLoS One. 2016 Dec 9;11(12):e0165616. doi: 10.1371/journal.pone.0165616. eCollection 2016.
Dust storms are devastating natural disasters that cost billions of dollars and many human lives every year. Using the Non-Hydrostatic Mesoscale Dust Model (NMM-dust), this research studies how different spatiotemporal resolutions of two input parameters (soil moisture and greenness vegetation fraction) impact the sensitivity and accuracy of a dust model. Experiments are conducted by simulating dust concentration during July 1-7, 2014, for the target area covering part of Arizona and California (31, 37, -118, -112), with a resolution of ~ 3 km. Using ground-based and satellite observations, this research validates the temporal evolution and spatial distribution of dust storm output from the NMM-dust, and quantifies model error using measurements of four evaluation metrics (mean bias error, root mean square error, correlation coefficient and fractional gross error). Results showed that the default configuration of NMM-dust (with a low spatiotemporal resolution of both input parameters) generates an overestimation of Aerosol Optical Depth (AOD). Although it is able to qualitatively reproduce the temporal trend of the dust event, the default configuration of NMM-dust cannot fully capture its actual spatial distribution. Adjusting the spatiotemporal resolution of soil moisture and vegetation cover datasets showed that the model is sensitive to both parameters. Increasing the spatiotemporal resolution of soil moisture effectively reduces model's overestimation of AOD, while increasing the spatiotemporal resolution of vegetation cover changes the spatial distribution of reproduced dust storm. The adjustment of both parameters enables NMM-dust to capture the spatial distribution of dust storms, as well as reproducing more accurate dust concentration.
沙尘暴是极具破坏力的自然灾害，每年造成数十亿美元的损失和许多人员伤亡。本研究使用非静力中尺度沙尘模型（NMM-dust），探究两个输入参数（土壤湿度和植被绿度分数）的不同时空分辨率如何影响沙尘模型的敏感性和准确性。通过模拟2014年7月1日至7日目标区域（覆盖亚利桑那州和加利福尼亚州部分地区，坐标为31, 37, -118, -112）的沙尘浓度进行实验，分辨率约为3千米。本研究利用地面和卫星观测数据，验证了NMM-dust沙尘风暴输出的时间演变和空间分布，并使用四个评估指标（平均偏差误差、均方根误差、相关系数和分数总误差）的测量值对模型误差进行了量化。结果表明，NMM-dust的默认配置（两个输入参数的时空分辨率均较低）会导致对气溶胶光学厚度（AOD）的高估。尽管它能够定性地再现沙尘事件的时间趋势，但NMM-dust的默认配置无法完全捕捉其实际空间分布。调整土壤湿度和植被覆盖数据集的时空分辨率表明，该模型对这两个参数都很敏感。提高土壤湿度的时空分辨率可有效降低模型对AOD的高估，而提高植被覆盖的时空分辨率则会改变再现沙尘风暴的空间分布。对这两个参数的调整使NMM-dust能够捕捉沙尘风暴的空间分布，并再现更准确的沙尘浓度。
