青藏高原小嵩草高寒草甸返青期遥感识别方法筛选
Identifying an optimal method for estimating green-up date of Kobresia pygmaea alpine meadow in Qinghai-Tibetan Plateau
- 2014年18卷第5期 页码:1117-1127
纸质出版日期: 2014
DOI: 10.11834/jrs.20143299
扫 描 看 全 文
浏览全部资源
扫码关注微信
纸质出版日期: 2014 ,
扫 描 看 全 文
[1]范德芹,朱文泉,潘耀忠,姜楠,郑周涛.青藏高原小嵩草高寒草甸返青期遥感识别方法筛选[J].遥感学报,2014,18(05):1117-1127.
FAN Deqin, ZHU Wenquan, PAN Yaozhong, et al. Identifying an optimal method for estimating green-up date of Kobresia pygmaea alpine meadow in Qinghai-Tibetan Plateau[J]. Journal of Remote Sensing, 2014,18(5):1117-1127.
小嵩草高寒草甸是青藏高原的主要植被类型
研究其返青期识别方法对于模拟及预测青藏高原植被物候变化具有重要意义。常用的植被返青期遥感识别方法主要是先对遥感植被指数原始时序数据进行拟合去噪声再求取返青期
各种方法对研究区域、研究经验、参数设置、函数初值设置等有很强的依赖性。为避免返青期识别方法在曲线拟合时对参数初值的依赖性和陷入局部最优解
本文引入了模拟退火算法对双高斯和双逻辑斯蒂函数进行参数优化
并分别对基于以上两种函数及多项式拟合的植被指数时序曲线进行对比
从而选出最佳拟合方法
最后采用最大斜率阈值法、动态阈值法和曲率法识别返青期。利用青藏高原小嵩草高寒草甸34个样本点的返青期地面观测数据及相应的8 km分辨率的NOAA归一化差值植被指数(NDVI)时序数据对以上各种组合的返青期遥感识别方案进行了测试
并选取了153个遥感实验点求取了近30年(1982年—2011年)青藏高原小嵩草高寒草甸的返青期
结果表明:采用双高斯函数拟合的NDVI曲线与原始NDVI时序数据最为接近
在此基础上采用最大斜率阈值法识别的小嵩草高寒草甸返青期及其变化趋势与地面物候观测结果最为一致;同时发现近30年青藏高原小嵩草高寒草甸的平均返青期主要集中在每年的第120—140天
并且呈逐年提前趋势
30年来提前了7天。
The Kobresia pygmaea alpine meadow is a main vegetation type in the Qinghai-Tibetan Plateau. An accurate detection of the green-up dates for K. pygmaea is important to simulate and predict vegetation phenology shifts under the influence of c limate change in the Qinghai-Tibetan Plateau. Green-up date estimation methods from remote sensing data generally include two processes: reconstruction of high-quality vegetation index time-series data through noise removal and calculation of green-up dates from the reconstructed vegetation index time series. The reconstruction methods for vegetation index time-series d ata can be divided into two categories: filter fitting and curve fitting methods. The green-up date retrieval methods include the threshold
maximum slope
curvature
and moving average methods. The green-up date identification method is a combination of the reconstruction methods for vegetation index time-series data and the retrieval methods for green-up dates under different study conditions. The a ccuracy of the green-up date identification methods is usually affected by many factors
such as specific geographic location
prior experience
parameterization
and initial parameters. In this study
we adopted a simulated annealing algorithm to optimize the r econstruction process and thus avoid the problems of low efficiency and local optimum caused by traditional optimal methods. We first used the double-Gaussian
double-Logistic
and polynomial functions to reconstruct the Normalized Difference Vegetation I ndex( NDVI)time series. After evaluations with visual inspections and root mean square error
we identified the most feasible r econstruction method. We then used the maximum slope
threshold
curvature
and dynamic threshold methods to derive the green-up dates from the best reconstructed NDVI time series. The performance of these three methods for green-up date identification were tested using the green-up data from 34 ground observation samples and their corresponding National Oceanic and A tmospheric Administration NDVI time-series data at 8-kilometer resolution. We selected additional 153 samples
which were e venly distributed in the K. pygmaea alpine meadow in the Qinghai-Tibetan Plateau
to test the identified optimal green-up estimation method and to investigate the changes in green-up dates in the study area. The reconstructed NDVI time series with the double-Gaussian function had the smallest deviation from the original NDVI time series
and the noises can be reduced effectively through the double-Gaussian fitting process.Therefore
the aforementioned method was the most suitable for describing the intra-annual growth cycle of the K. pygmaea alpine meadow. The reconstructed NDVI time series with the double-Gaussian function method indicated that the green-up dates identified with the maximum slope threshold method agreed with the observed ground phenology data. The correlation coefficients between the identified green-up dates and the observed dates were 0. 823( P < 0. 001) and 0. 646( P < 0. 01) at the Haiyan and Gande stations
respectively. The average green-up dates for the K. pygmaea alpine meadow in the Qinghai-Tibetan Plateau were mainly located between DOY( Day of Year) 120( i. e.
30 April) and 140( i. e.
20 May). The green-up onset date advanced by an average of 7 days from 1982 to 2011 in the study region.
小嵩草高寒草甸NDVI返青期模拟退火算法
Kobresia pygmaeaalpine meadowNDVIgreen-up datesimulated annealing algorithm
相关作者
相关机构