基于MODIS-LAI数据的广西甘蔗物候期提取

广西是我国甘蔗最大的种植区和蔗糖产业基地,但多年来因自然灾害导致的产量损失严重。及时、准确获取甘蔗的物候动态,可为区域甘蔗种植结构优化、灾害风险管理等提供科学支撑。本文分别借助Savitzky-Golay滤波、Asymmetric Gauss拟合及Double Logistic拟合方法对广西2014—2018年的甘蔗MODIS-LAI数据进行时间序列重构,并采用动态阈值法识别甘蔗关键物候期的时空变化特征。结果表明:1)上述方法可有效消除广西甘蔗LAI序列的不稳定波动和奇异值,且Asymmetric Gauss和Double Logistic方法的拟合结果较好地弥补了S-G滤波处理中局部曲线值动荡变化的缺点,能更准确地识别甘蔗生长过程;2)三种方法对广西甘蔗的播种-萌芽期、茎伸期、成熟期开始时间识别的RMSE和平均误差均在±15 d内,其中Asymmetric Gauss方法对甘蔗物候期的提取结果最优。综上,时序曲线重构法的抗干扰性能较好,可一定程度上弥补因云雨、复杂地形地貌导致LAI序列的误差,保证广西甘蔗物候信息获取的准确性。

Extraction of the phenology of sugarcane in Guangxi based on MODIS-LAI data

Guangxi is the largest sugarcane planting area and sucrose industry base in China, but its production has been seriously lost due to natural disasters for years. Therefore, timely and accurate acquisition of phenological dynamics of sugarcane can provide scientific support for the optimization of regional sugarcane planting structure, and disaster risk management. In this paper, the time-series reconstruction of MODIS-LAI data of inversed sugarcane in Guangxi from 2014 to 2018 was carried out by Savitzky-Golay filtering, Asymmetric Gauss fitting, and Double Logistic fitting methods, respectively. The spatio-temporal variation characteristics of key phenological features of sugarcane were identified by the dynamic threshold method. The results showed that: 1) The above mentioned methods could effectively eliminate unstable fluctuations and singular values of time-series LAI of sugarcane in Guangxi. The Asymmetric Gauss and Double fitting results of logistical method made up the disadvantage of local curve value fluctuation in S-G filtering process and could more accurately identify the growth processes of sugarcane; 2) The RMSE and average error for identification of the seeding-germination, stem elongation and the beginning of maturity period of sugarcane in Guangxi by the proposed three methods were within ±15d, whereas Asymmetric Gauss method was the best method to extract phenological period of sugarcane. This study indicated that the time series curve reconstruction method has good anti-interference performance, which can make up for the time-series LAI caused by cloudy, rainy, and complex terrain conditions. Our proposed method, to a certain extent, and ensure the accuracy of sugarcane phenological inversion in Guangxi.