基于Sentinel-1 SAR数据的冬小麦晚霜冻害等级监测方法研究

Sentinel-1 SAR-based Method for Monitoring Spring Frost Damage Severity and Levels in Winter Wheat

摘要:

晚霜冻害是影响冬小麦生产的重要农业气象灾害,会引发穗部受损并导致产量下降,因此,发展高效的监测方法对防灾减灾具有重要意义。本文以河南省滑县为研究区,针对2018年4月5—7日发生的冬小麦晚霜冻害开展等级监测研究。基于2018年1—6月的Sentinel-1合成孔径雷达(Synthetic aperture radar, SAR)影像,构建冻害严重度指数(Frost damage severity index, FDSI),用于识别冬小麦的冻害程度。采用51×51像元窗口的均值聚合方法对垂直发射-垂直接收(Vertical-vertical, VV)和垂直发射-水平接收(Vertical-horizontal, VH)波段进行降噪处理,在满足变异系数(Coefficient of variation, CV)和信噪比(Signal-to-noise ratio, SNR)稳定性的同时,有效平衡了噪声抑制与空间细节保留。基于VH/VV极化比(VH/VV polarization ratio, VH/VV)、雷达植被指数(Radar vegetationindex, RVI)与双极化SAR 植被指数(Dual polarization SAR vegetation index, DPSVI)等多个雷达指数,结合Savitzky-Golay(SG)滤波与包络线拟合方法,提取由霜冻事件引起的时序信号突降特征,并对突降幅度进行标准化处理,最终构建FDSI。实地调查样点的验证结果表明,基于VH/VV的FDSI与冻害等级之间的相关系数r达到0.82(p<0.001),显著优于RVI(r=0.76)和DPSVI(r=0.73)。研究区晚霜冻害空间制图显示,本次霜冻事件影响范围广,但以轻、中度冻害为主(61.8%),县域南部受灾程度最重。本文依据冬小麦晚霜冻害的地面分级标准,构建了基于遥感数据的监测方法,可为区域尺度的标准化冻害监测提供高效、可靠的技术方案。

Abstract:

Spring frost damage is a major agricultural meteorological disaster affecting the production of winter wheat, which can cause damage to the ear parts and lead to a decrease in yield. Therefore, developing an efficient monitoring method is of great significance for disaster prevention and mitigation. Focusing on Hua County, Henan Province, where a spring frost event occurred from April 5th to 7th, 2018, the research on monitoring spring frost damage severity and levels in winter wheat was conducted. Sentinel-1 synthetic aperture radar (SAR) images from January to June 2018 were used to develop a frost damage severity index (FDSI) for identifying the severity of frost damage in winter wheat. A mean aggregation method with a 51×51 pixel window was adopted to conduct noise reduction processing on the vertical-vertical (VV) and vertical-horizontal (VH) bands. While satisfying the stability of the coefficient of variation (CV) and signal-to-noise ratio (SNR), it effectively balanced noise suppression and spatial detail preservation. Based on multiple radar indices, including the VH/VV polarization ratio (VH/VV), radar vegetation index (RVI), and dual polarization SAR vegetation index (DPSVI), combined with the Savitzky-Golay (SG) filtering and envelope fitting techniques, the temporal signal drop characteristics caused by the frost event were extracted. The drop amplitude was standardized to finally construct the FDSI. The validation results from field survey sampling points showed that the correlation coefficient between the FDSI based on VH/VV and the frost damage level reached 0.82 (p<0.001), which was significantly better than RVI (r=0.76) and DPSVI (r=0.73). The spatial mapping of spring frost damage showed that the frost event affected a broad area, but the damage was mainly light to moderate (61.8% ), with the southern part of the county being the most severely impacted. A remote sensing-based monitoring method for spring frost damage in winter wheat was developed. Utilizing the ground-based damage level classification standard, the method can provide an efficient and reliable approach for standardized monitoring at a regional scale.