基于MODIS NDVI时序数据的华北平原双季作物物候特征遥感监测研究

Remote Sensing Monitoring of Phenological Characteristics of Double-Cropping Crops in the North China Plain Based on MODIS NDVI Time-Series Data

摘要: 在精细化监测大尺度农作物物候节律对于评估区域粮食安全和优化农田管理具有重要意义。本研究基于NASA AppEEARS平台获取了华北平原核心农田区(114˚E~120˚E, 32˚N~40˚N)高时空分辨率的MOD13A2 (Version 061)逐16天1 km网格NDVI遥感产品。利用MATLAB对连续两个农业年(共47期)的时序数据进行空间边界严格裁剪、掩膜与区域均值统计重建，排除了复杂山地与大面积水体的混合像元噪声，成功提取了高纯净度的区域作物物候长势时序曲线。结果表明：华北平原核心农田区的NDVI时序曲线呈现出极其典型且规律的“双峰一谷”年内物候特征，完美映射了该区域“冬小麦–夏玉米”一年两熟制的典型农业轮作景观。在完整的一年轮作周期内，曲线精准捕捉到了关键的农业物候节点。春季4至5月间(时序第9期与第32期)，冬小麦进入抽穗灌浆盛期，形成年内第一个NDVI次高峰(均值约为0.56~0.58)；随后在6月份，随着冬小麦的集中机械化收割，地表大面积裸露，NDVI曲线发生断崖式下跌，在第12期与第35期形成显著的麦收“谷底”(均值降至0.35左右)。7月以后，播种后的夏玉米在高温多雨驱动下快速生长，于8至9月间(对应时序第16期与第39期)达到全年的绝对波峰(均值超过0.72)，显现出夏玉米盛期的生物量积累与群落冠层茂密程度显著高于春季小麦盛期。两个完整农业年内双峰特征在时间位相与数值量级上表现出高度的年际稳定性和可重复性。本研究建立的数据处理与物候提取方案能够有效剔除外界随机噪声，获取高纯净度的陆面物候背景场数据，可为后续开展高级遥感参数反演、作物产量预测以及区域生态数值模拟验证提供规范、坚实的数据支撑。

Abstract: A Detailed monitoring of large-scale crop phenological rhythms is of great significance for assessing regional food security and optimizing farmland management. In this study, high-spatiotemporal-resolution MOD13A2 (Version 061) 16-day 1 km grid NDVI remote sensing products were obtained from the NASA AppEEARS platform for the core agricultural area of the North China Plain (114˚E~120˚E, 32˚N~40˚N). Using MATLAB, the time-series data from two consecutive agricultural years (a total of 47 periods) underwent strict spatial boundary cropping, masking, and regional mean statistical reconstruction. This process eliminated mixed-pixel noise from complex mountainous terrain and large water bodies, successfully extracting high-purity regional crop phenological growth time-series curves. The results indicate that the NDVI time series curves in the core farmland areas of the North China Plain exhibit an extremely typical and regular “two peaks and one trough” annual phenological pattern, perfectly reflecting the region’s typical “winter wheat–summer maize” two-crop-per-year agricultural rotation landscape. Throughout the complete annual rotation cycle, the curves accurately captured key agricultural phenological milestones. Between April and May (periods 9 and 32), winter wheat enters the peak stage of heading and grain filling, forming the first annual NDVI sub-peak (with an average of approximately 0.56~0.58); Subsequently, in June, as winter wheat was harvested en masse using mechanized methods, large areas of the ground were exposed, causing the NDVI curve to plummet precipitously, forming a distinct “trough” during the wheat harvest in periods 12 and 35 (with the mean dropping to around 0.35). After July, newly planted summer corn grew rapidly under the influence of high temperatures and abundant rainfall, reaching its annual absolute peak between August and September (corresponding to periods 16 and 39), with an average value exceeding 0.72. This indicates that biomass accumulation and canopy density during the peak growth period of summer corn were significantly higher than those of spring wheat. The bimodal pattern observed over two complete agricultural years exhibited high interannual stability and reproducibility in both temporal phase and magnitude. The data processing and phenological extraction protocol established in this study effectively filters out external random noise to obtain high-purity land surface phenological background data, providing standardized and robust data support for subsequent advanced remote sensing parameter inversion, crop yield forecasting, and validation of regional ecological numerical simulations.