Phenomapping of rangelands in South Africa using time series of RapidEye data
Phenomapping of rangelands in South Africa using time series of RapidEye data
复制标题
DOI：
10.1016/j.jag.2016.08.001
复制
发表时间：
2016-12-01
影响因子：
7.5
通讯作者：
Schellberg, Juergen
中科院分区：
文献类型：
--
作者：
Parplies, Andre;Dubovyk, Olena;Schellberg, Juergen
物候制图是一种基于植被指数时间序列推导植被物候和草地生产力空间格局的方法。在这项研究中，我们提出了一种新的空间制图方法，该方法将高分辨率（HR）卫星时间序列的物候计量学与空间逻辑回归模型相结合，以区分牧场土地管理系统。从RapidEye时间序列中选取南非的牧场，我们计算了两周降噪归一化植被指数（NDVI）图像。对于2011-2012年生长季，我们进一步推导了生长季开始、结束和长度等主要物候指标以及NDVI曲线的振幅、左导数和小积分等相关物候变量。然后，我们以非常详细的5米空间分辨率绘制了两种不同权属系统（公共和商业）的这些物候特征图。二元逻辑回归（BLR）的结果表明，NDVI曲线的振幅和左导数具有统计学意义。这些指标有助于区分商业牧场系统和公共牧场系统。我们的结论是，物候图与空间建模相结合是一种强大的工具，可以有效地聚集物候和生产力指标，从而对作物物候与现场条件和管理的关系进行空间明确分析。这种方法特别有可能用于分散和不完整的环境，例如半干旱的南非的农业系统，那里的物候在不同年份之间和年内变化很大。此外，我们看到了一个强有力的视角现象映射，以支持空间明确的植被建模。(C) 2016 Elsevier B.V.版权所有
Phenomapping is an approach which allows the derivation of spatial patterns of vegetation phenology and rangeland productivity based on time series of vegetation indices. In our study, we propose a new spatial mapping approach which combines phenometrics derived from high resolution (HR) satellite time series with spatial logistic regression modeling to discriminate land management systems in rangelands. From the RapidEye time series for selected rangelands in South Africa, we calculated bi-weekly noise reduced Normalized Difference Vegetation Index (NDVI) images. For the growing season of 2011-2012, we further derived principal phenology metrics such as start, end and length of growing season and related phenological variables such as amplitude, left derivative and small integral of the NDVI curve. We then mapped these phenometrics across two different tenure systems, communal and commercial, at the very detailed spatial resolution of 5 m. The result of a binary logistic regression (BLR) has shown that the amplitude and the left derivative of the NDVI curve were statistically significant. These indicators are useful to discriminate commercial from communal rangeland systems. We conclude that phenomapping combined with spatial modeling is a powerful tool that allows efficient aggregation of phenology and productivity metrics for spatially explicit analysis of the relationships of crop phenology with site conditions and management. This approach has particular potential for disaggregated and patchy environments such as in farming systems in semi-arid South Africa, where phenology varies considerably among and within years. Further, we see a strong perspective for phenomapping to support spatially explicit modelling of vegetation. (C) 2016 Elsevier B.V. All rights reserved.