基于遥感与模型耦合的冬小麦生长预测

遥感的空间性、实时性与作物生长模型的过程性、机理性优势互补,将两者有效耦合已成为提高作物生长监测预测能力的重要手段之一。提出了一种基于地空遥感信息与生长模型耦合的冬小麦预测方法,该方法基于初始化/参数化策略,以不同生育时期的小麦叶面积指数(LAI)和叶片氮积累量(LNA)为信息融合点将地面光谱数据(ASD)及HJ-1 A/B CCD、Landsat-5 TM数据与冬小麦生长模型(WheatGrow)耦合,反演得到区域尺度生长模型运行时难以准确获取的部分管理措施参数(播种期、播种量和施氮量),在此基础上实现了对冬小麦生长的有效预测。实例分析结果表明,LNA较LAI对模型更敏感,以之作为耦合点的反演效果较好。另外,抽穗期是遥感信息与生长模型耦合的最佳时机,对播种期、播种量和施氮量反演的RMSE值分别达到5.32 d、14.81 kg/hm²、14.11 kg/hm²。生长模型与遥感耦合后的模拟结果很好地描述了冬小麦长势和生产力指标的时空分布状况,长势指标的模拟相对误差小于0.25,籽粒产量模拟的相对误差小于0.1。因此研究结果可为区域尺度冬小麦生长的监测预测提供重要理论依据。

Predicting winter wheat growth based on integrating remote sensing and crop growth modeling techniques

Historically, remote sensing (RS) and crop growth modeling have been used independently to monitor and predict crop growth. This paper reports a growth prediction technique for winter wheat (Triticum aestivum L.) based on the integration of ground-based and space-borne remote sensing data and a winter wheat growth model (WheatGrow). Leaf area index (LAI) and leaf nitrogen accumulation (LNA) of winter wheat were estimated using ASD field spectrometer, HJ-1 A/B CCD, and Landsat-5 TM data and statistical remote sensing estimation models. This information was integrated with the WheatGrow model in three different growth stages (jointing, heading, and grain filling). Management parameters included sowing date, sowing rate, and nitrogen rate. Parameterization for regionalization of the integrated model was accomplished using the Shuffled Complex Evolution-University of Arizona (SCE-UA) optimization algorithm. This integrated technique was tested on independent datasets acquired from three winter wheat field tests in different years on different winter wheat varieties and at different treatments of nitrogen rates and sowing densities, and from data obtained from study areas in Hai'an and Rugao counties in Jiangsu Province (in central eastern China), both of which are main production areas of high-quality, low-gluten wheat in China. The results showed that LNA, one of the most sensitive parameters within WheatGrow, was better than LAI as an integrated parameter for crop model parameter initialization with the best integration period being the heading stage. RMSE values were 5.32 days, 14.81 kg/hm² and 14.11 kg/hm² for sowing date, sowing rate, and nitrogen rate based on the ground spectral datasets, and 6.55 days, 13.94 kg/hm² and 84.97 kg/hm² based on the space-borne satellite images. The crop model parameterization results were poorest when the grain filling stage was used as the integration period, probably because crop growth at earlier stages would be more influenced by agronomic management measures. In addition, predicted results well described the temporal and spatial distribution of winter wheat growth status and productivity in the study area, with relative error values of 0.13, 0.18 and 0.03 for LAI, LNA and grain yield based on the ground spectral datasets, and 0.22, 0.23 and 0.06 based on the satellite images. The error may be due to the limited simulation ability of the WheatGrow model, or it may have been generated from the process of remote sensing information extraction and the statistical remote sensing estimating models, all of which need improvement. Nevertheless, the study has provided an important step toward more routine use of using remote sensing and crop modeling techniques together to improve our ability of regional monitoring and yield prediction of winter wheat.