植被聚集指数（clumping index ，CI）是用于定义植被冠层聚集效应——冠层叶片在空间上的非随机分布现象的参数。CI作为植被冠层间隙率模型中的一个参数，其大小变化能直接影响某些遥感中间参数，如有效叶面积指数（LAIe）的计算，也能进一步影响陆地植被生物量、陆地生态系统碳循环以及其他相关遥感应用的计算结果。以往的研究中一般假设CI是一个固定值，但理论上CI并非一个静态参数，其大小应当会随着植被物候变化而变化，这种变化往往具有季节性。目前对于如何提取计算CI的研究较多，但极少有对CI季节性变化的专门研究。本文尝试使用快速傅里叶变换（FFT）的方法，基于全球2001-2019年逐月500m分辨率的MODIS数据计算获取的CI遥感产品，对全球CI季节性变化展开初步研究。对数据预处理使用3*3大小的窗口对数据升尺度到1500*1500从而适度缩减计算量并平滑空间上的尖锐噪声，记录下9个小像元的CI标准差（std）作为空间异质性标准，经过直方图统计选取0.3作为std阈值，排除掉所有std大于0.3的像元。接着逐像元记录数据缺失比例，使用三次样条插值填补部分像元时间序列上的空缺值，从而得到完整时间序列，并使用窗口长度3的均值滤波再次平滑时间序列散点曲线上的尖锐噪声；同时，以时间序列上缺失值占比不小于90%及std<0.3的数量占比不小于90%为标准筛选合格像元，在此基础上，每个合格像元均以原始数据的质量（QA）众数作为合格像元的新质量指标参数nQA（不合格像元nQA记为0）；预处理完成后通过逐像元的快速傅里叶变换，将时间序列散点曲线分解为一系列不同周期，振幅，初相位的正弦波的组合，选取其中振幅最大的正弦波散点曲线，记录其波形参数（最大振幅mA，最大振幅周期mT，初相位p）；其后逐像元计算19年来CI均值（mean），并使用最大振幅mA的两倍除以mean计算获得每个像元的振幅占比（sinf）。根据模拟的正弦波公式，对最大振幅周器mT为12（月）的像元的初相位p进行变换计算获取该像元在一年（12个月）内CI达到高峰值和低峰值的月份。结果表明：（1）在nQA>0的像元中，sinf分布范围峰值出现在5%-15%，特别在北美洲中西部地区高达25%以上，这表明CI值的变化波动不可被简单忽略。（2）在nQA>0的像元中，mT众数为12（月，即1年），占比达到76%，远远大于mT其他任何单一值占比，说明全球范围内广泛存在着CI的季节性变化波动。(3)对mT=12（月）的像元计算了CI理论达到低峰值月份，全球CI的低峰值出现月份主要在冬，夏两季，北半球以6,7月份居多，南半球以12,1月份居多。且CI低峰值出现月份的分布大体呈现出纬度地带性。对于某些地中海气候区，CI低峰值出现月份表现出反季节性，与当地雨季（冬季）保持一致。（4）合格像元（nQA>0）主要分布于温带地区。热带及亚热带部分地区的数据在计算过程中被大量“筛除”，导致本研究结论实际针对的地区主要是温带区域。这可能是由于多云天气导致数据缺失过多造成的。若要对热带亚热带地区进行类似分析，后续可考虑适度调整筛选阈值，并适当结合地面数据进行分析。

Clumping index (CI) is a parameter used to define the clustering effect of vegetation canopy. As there are various organizational structures in the vegetation canopy on the earth’s surface, its leaves are not randomly distributed in space, which is called agglomeration effect. CI is defined in the gap rate model of vegetation canopy to quantitatively describe the non-random distribution of leaves, which is often used to model vegetation. Therefore, CI is of great significance to the study of vegetation biomass, terrestrial ecosystem carbon cycle and other related remote sensing applications. However, CI is not a long-term static parameter. Due to the phenology of vegetation, CI, which reflects the aggregation of vegetation leaves, should also have some seasonal changes accordingly, which will directly affect the results of any surface modeling considering CI. However, previous studies have generally assumed that CI is a fixed value. While there has been many studies on how to extract and calculate CI, and some reliable long-time series global CI products have been generated by several methods, there are few systematic studies on the seasonal change of CI. Based on the monthly CI data in recent 19 years, this research studies the seasonal changes of CI in the world. The main research methods are as follows: scaling the data in a 3*3 window and calculating the corres-ponding standard deviation (std), filling in the vacancy values by cubic spline interpolation, calculating the monthly average of 19-year data, and filtering the data with one-dimensional smooth average to obtain the complete CI time series of each pixel. Then a new quality index, nQA, which considers the time integrity, spatial heterogeneity and original data quality QA, is calculated. nQA > 0 indicates that the missing ratio of the pixel data is less than 0.1, and the std value of more than 90% of the pixel data is less than 0.3. One-dimensional Fourier filtering method is used to decompose the obtained sequence to obtain a series of cosine waves. Taking the month as the basic time unit, the corresponding period (maximum amplitude period, mT) of the wave with the largest amplitude is taken and recorded, and the pixels with mT=12 (month) are further analyzed. For the pixel with mT=12, record the maximum amplitude of the corresponding cosine wave, record the initial phase and calculate the theoretical month in which CI reaches its peak value. Record the ratio of twice the maximum amplitude to the average value (amplitude ratio, sinf) as an index to indicate the seasonal fluctuation intensity of CI. The results show that the CI of most regions with nQA>0 in the world fluctuates seasonally with a period of 12 months, as mT=12 accounting for 76.22%. And the peak of sinf probability density is 5%-15% in the whole world, especially over 25% in the midwest of North America, which indicate that the impact of seasonal fluctuations can not be ignored. The results of the low peak months of CI obtained from the initial phase show that the low peak months of CI in the world are mainly in winter and summer, with June and July in the northern hemisphere and December and January in the southern hemisphere. And the distribution of the low peak of CI in the month is roughly parallel to the latitude, showing latitude zonality. For some Mediterranean climate regions, such as northern Africa and southern Australia, the month of low CI peak shows anti-seasonality, which is consistent with the local rainy season (i.e winter for Mediterranean climate regions).