农区MODIS植被指数时间序列数据重建

Reconstructing MODIS vegetation index time-series data of cropping area

MODIS植被指数时间序列数据能够连续反映植被的覆盖情况,是农作物遥感测量的重要数据源。但现有MOD13产品中存在由云、气溶胶,传感器角度等干扰因素导致的噪声指数。因此,必须对MOD13时间序列中的噪声指数进行恢复。根据农作物种植区物候与熟制信息,将待重建像元时间序列划分为符合作物生长周期的时段。对各时段内指数按非对称高斯模型重建,优化相邻时段之间重叠期内指数。多次迭代重建和优化过程后恢复时间序列中噪声指数。对覆盖北京市通州区以南和河北省保定市以北农区2005年MOD13数据进行重建,与两阶段S-G滤波重建结果对比。结果表明:噪声指数被准确判断并恢复。农区多熟制导致的低值指数被有效保留。重建时间序列可以正确反映植被的覆盖情况。

As an excellent dataset which reflects the coverage of vegetation on earth surface continuously, MODIS vegetation index time series have already become an important data source in crop measurement by remote sensing. However, there are always some noises caused by atmosphere variability and sensor angle in MOD13 vegetation product For this reason, the time series of MOD13 should be reconstructed before application. According to phenology and cropping system, time series were separated into different periods related to the growth process of crops. Then, the vegetation indices in each period were reconstructed based on asymmetric gaussian function. After all periods were reconstructed, the indices in overlapping range between two adjacent periods were optimized. The above two procedures were repeated a certain times to restore the indices affected by noises. The proposed method were applied to reconstruct the NDVI time series of cropping area lies between Tong zhou District, Beijing, and Baoding City, HeBei Province with MOD13 data acquired in 2005. The same data were reconstructed by two step Savitzky-Golay filter. The comparison between two results show that the noise in time series can be evaluated and restored accurately. Meantime, the low vegetation indices caused by double cropping system are reserved effectively. The whole reconstructed NDVI time series can indicate vegetation coverage accurately.