Research in Spatial Ecology : Mathematical Studies of Spatio-Temporal Dynamics

近年来，空间生态数据的积累主要来自于长期生态研究点或远程观测技术。土地利用和土地覆盖的空间格局变化以一系列地图的形式给出。在本研究项目中，笔者基于空间马尔可夫模型，提出了一种新的基于时空数据的生态系统过程和人类影响估算的数学方法。通过这种方法，我们可以得到一个生态系统变化的基本理论框架从地图数据估计生态过程。自20世纪80年代初以来，对一个50公顷的新热带森林进行了大面积普查。我们与Steve Hubbell和他的同事合作，分析了2年来植被高度空间数据的时间序列。50公顷的地块被划分为约2000个地块，每个地块面积为5 × 5米。我根据最大植被高度将立地分为林隙和非林隙。然后建立了描述空间格局动态的马尔可夫过渡模型。结果表明，林冠样地向林隙样地的过渡速率和林隙样地的恢复速率与林隙样地的过渡速率密切相关。因此，空间格局将变得非随机，间隙点彼此聚集。我们还认识到，简单的估计过渡率参数的方法包含很大的偏差。这是因为一个站点的转换会改变其相邻站点的转换速率。为了消除偏见，我们使用了一种名为NCBC的方法，这是由Hakoyama博士和我自己开发的基于土地利用变化的建模。将样地分为森林、农地、道路等，然后对这些离散状态之间的过渡过程进行建模。从空间数据中，我们可以知道邻域对转换速率的影响。我们开发了一个马尔可夫空间模型，可以根据土地利用变化预测森林损失的未来变化。

Recently, spatial ecological data have been accumulated from long-term ecological research sites or from remote censing techniques. The change in the spatial pattern of land use and land cover given as a series of maps. In this research project, I develop a new mathematical method to estimate the ecosystem process and the human impact from spatio-temporal data, based on spatial Markov model. By this way we might obtain a basic theoretical framework of ecosystem change.[1] Estimating ecological process from map data.BCI 50ha plot, large-area census of a neotropical forest has been carried out since early 1980's. In collaboration with Steve Hubbell and his colleagues, we analyzed times series of vegetation height spatial data over 2 years. The 50ha plot was separated into about 2000 sites each having 5x5m. I classified sites to gaps and non-gaps according to the maximum vegetation height. Then developed Markovian transition model describing the dynamics of spatial patterns. We observed that the transition rate from a canopy site to a gap site and the recovery rate (the opposite transition) depend strongly on the neighbors. As a consequence, the spatial pattern would become non-random with gap sites aggregated with each other.We also recognized that the simple way of estimating transition rate parameter includes a large bias. This is because the transition at a site would change the rate of transition in its neighbors. To remove the bias we used a method named NCBC, which was developed by Dr Hakoyama and myself.[2] Modeling based on land use change.Sites are classified to forest, agricultural land, road etc., and then model the process of transition between these discrete states. From spatial data, we can know the effect of neighbors to modify the transition rate. We developed a Markovian spatial model that can predict future change in the loss of forests based on land use change.