Prasad V Krishna, Anuradha E, Badarinath K V S
Forestry and Ecology Division, National Remote Sensing Agency (Department of Space--Government of India), India.
Int J Biometeorol. 2005 Sep;50(1):6-16. doi: 10.1007/s00484-005-0268-0. Epub 2005 May 18.
Ten-day advanced very high resolution radiometer images from 1990 to 2000 were used to examine spatial patterns in the normalized difference vegetation index (NDVI) and their relationships with climatic variables for four contrasting forest types in India. The NDVI signal has been extracted from homogeneous vegetation patches and has been found to be distinct for deciduous and evergreen forest types, although the mixed-deciduous signal was close to the deciduous ones. To examine the decadal response of the satellite-measured vegetation phenology to climate variability, seven different NDVI metrics were calculated using the 11-year NDVI data. Results suggested strong spatial variability in forest NDVI metrics. Among the forest types studied, wet evergreen forests of north-east India had highest mean NDVI (0.692) followed by evergreen forests of the Western Ghats (0.529), mixed deciduous forests (0.519) and finally dry deciduous forests (0.421). The sum of NDVI (SNDVI) and the time-integrated NDVI followed a similar pattern, although the values for mixed deciduous forests were closer to those for evergreen forests of the Western Ghats. Dry deciduous forests had higher values of inter-annual range (RNDVI) and low mean NDVI, also coinciding with a high SD and thus a high coefficient of variation (CV) in NDVI (CVNDVI). SNDVI has been found to be high for wet evergreen forests of north-east India, followed by evergreen forests of the Western Ghats, mixed deciduous forests and dry deciduous forests. Further, the maximum NDVI values of wet evergreen forests of north-east India (0.624) coincided with relatively high annual total precipitation (2,238.9 mm). The time lags had a strong influence in the correlation coefficients between annual total rainfall and NDVI. The correlation coefficients were found to be comparatively high (R2=0.635) for dry deciduous forests than for evergreen forests and mixed deciduous forests, when the precipitation data with a lag of 30 days was correlated against NDVI. Using multiple regression approach models were developed for individual forest types using 16 different climatic indices. A high proportion of the temporal variance (>90%) has been accounted for by three of the precipitation parameters (maximum precipitation, precipitation of the wettest quarter and driest quarter) and two of the temperature parameters (annual mean temperature and temperature of the coldest quarter) for mixed deciduous forests. Similarly, in the case of deciduous forests, four precipitation parameters and three temperature parameters explained nearly 83.6% of the variance. These results suggest differences in the relationship between NDVI and climatic variables based upon the time of growing season, time interval and climatic indices over which they were summed. These results have implications for forest cover mapping and monitoring in tropical regions of India.
利用1990年至2000年的十天期先进甚高分辨率辐射计图像，研究了印度四种不同森林类型的归一化植被指数（NDVI）的空间格局及其与气候变量的关系。NDVI信号已从均匀植被斑块中提取出来，发现落叶林和常绿林类型的信号明显不同，尽管混交落叶林的信号与落叶林的信号相近。为了研究卫星测量的植被物候对气候变化的十年响应，利用11年的NDVI数据计算了7种不同的NDVI指标。结果表明，森林NDVI指标存在很强的空间变异性。在所研究的森林类型中，印度东北部的湿润常绿林平均NDVI最高（0.692），其次是西高止山脉的常绿林（0.529）、混交落叶林（0.519），最后是干燥落叶林（0.421）。NDVI总和（SNDVI）和时间积分NDVI遵循类似的模式，尽管混交落叶林的值更接近西高止山脉常绿林的值。干燥落叶林的年际变化范围（RNDVI）值较高，平均NDVI较低，同时标准差也较高，因此NDVI的变异系数（CVNDVI）也较高。已发现印度东北部湿润常绿林的SNDVI较高，其次是西高止山脉的常绿林、混交落叶林和干燥落叶林。此外，印度东北部湿润常绿林的最大NDVI值（0.624）与相对较高的年总降水量（2238.9毫米）相吻合。时间滞后对年总降水量与NDVI之间的相关系数有很大影响。当将滞后30天的降水数据与NDVI进行相关分析时，发现干燥落叶林的相关系数（R2 = 0.635）比常绿林和混交落叶林的相关系数更高。使用多元回归方法，利用16种不同的气候指数为各森林类型建立了模型。对于混交落叶林，三个降水参数（最大降水量、最湿润季度降水量和最干燥季度降水量）和两个温度参数（年平均温度和最寒冷季度温度）解释了大部分时间方差（>90%）。同样，对于落叶林，四个降水参数和三个温度参数解释了近83.6%的方差。这些结果表明，基于生长季节的时间、时间间隔以及求和的气候指数，NDVI与气候变量之间的关系存在差异。这些结果对印度热带地区的森林覆盖制图和监测具有重要意义。
