全球1km5天分辨率NDVI数据集（2010-2020）

全球1km分辨率NDVI数据集是在5种1km分辨率卫星遥感数据（Terra/MODIS、Aqua/MODIS、NOAA18/AVHRR、FY3A/VIRR、FY3B/VIRR）基础上，参照标准的MODIS-NDVI合成算法生产的5-day分辨率的数据集。本数据集的合成算法分两步：首先，在5-day合成期内，利用基于带NDVI权重的核系数稳健拟合法，把多源数据按照残差阈值10%、20%和大于20%划分为三个质量等级，其中残差阈值大于20%的三级数据质量最差，不参与合成。然后，根据5-day合成期内各质量等级的有效观测数目，分别采取主算法或备用算法进行NDVI合成，获得了2015-2020年全球时间分辨率为5-day、空间分辨率为1km的NDVI时间序列。在中国黑河流域利用高分辨率ASTER影像，将数据集与标准MODIS NDVI产品进行对比分析。结果表明，基于多源数据集生成的NDVI，与ASTER-NDVI的决定系数R2显著高于基于单源数据集生成的标准MODIS-NDVI。该数据集采用Sinusoidal Tile Grid分幅，每景覆盖的经纬度范围为10°x10°，该数据集存储为*.tif 格式。

The global 1-kilometer spatial resolution NDVI dataset is a 5-day temporal resolution dataset produced based on five 1-kilometer spatial resolution satellite remote sensing datasets (Terra/MODIS, Aqua/MODIS, NOAA18/AVHRR, FY3A/VIRR, FY3B/VIRR) following the standard MODIS-NDVI synthesis algorithm.

The synthesis algorithm of this dataset consists of two steps: First, during the 5-day synthesis period, multi-source data are classified into three quality grades based on residual thresholds of 10%, 20%, and greater than 20% using a robust kernel coefficient fitting method with NDVI weighting. The third-grade data with residual threshold exceeding 20% have the poorest quality and are excluded from the synthesis. Then, according to the number of valid observations of each quality grade during the 5-day synthesis period, the primary or backup algorithm is respectively applied for NDVI synthesis, generating global NDVI time series from 2015 to 2020 with a temporal resolution of 5 days and spatial resolution of 1 kilometer.

A high-resolution ASTER image in the Heihe River Basin of China was used to perform a comparative analysis between this dataset and the standard MODIS NDVI product. The results indicate that the coefficient of determination R² between the NDVI derived from the multi-source dataset and ASTER-NDVI is significantly higher than that of the standard MODIS-NDVI generated from single-source datasets.

This dataset adopts the Sinusoidal Tile Grid tiling scheme, where each tile covers a 10°×10° latitude-longitude range, and is stored in *.tif format.

搜集汇总

背景与挑战

背景概述

该数据集是一个全球范围的NDVI（归一化植被指数）数据集，覆盖2010年至2020年，具有1km空间分辨率和5天时间分辨率。它基于多源卫星遥感数据（包括Terra/MODIS、Aqua/MODIS等），采用改进的合成算法，通过质量分级和有效观测数量优化NDVI合成，相比单源标准MODIS产品，在验证中显示出更高的精度（决定系数R2更高）。数据集以Sinusoidal投影存储为.tif格式，但目前未开放下载。