Data: Probabilistic Cellular Automata Modelling and Simulation of Land Use Changes in Okomu National Park

This study monitors and models spatio–temporal land use changes in Okomu National Park over two decades (2000 – 2020) with the aim of projecting forest cover changes for the near future. A probabilistic cellular automata (CA) model was created and used to simulate land use changes with the aim of prediction. Landsat 7 ETM+ satellite images for years 2000, 2005, 2010, 2015, and 2020 were classified into Forest and Non–Forest using maximum likelihood supervised classification algorithm. The overall classification accuracy for the years under study was 98.1838%, 97.5169%, 96.3325%, 91.6647%, and 94.6124% with overall kappa coefficients of 0.9654, 0.9557, 0.9524, 0.8563, and 0.9094 respectively. State transition probabilities for 2000–2005, 2005–2010, 2010–2015, and 2015–2020 were calculated from the classified images. A probabilistic cellular automata model using Moore’s neighborhood with a Von Neumann extension was used to simulate land use changes for years 2005, 2010, 2015 and 2020 with year 2000 as the base year. Simulation accuracy was 77.46% for year 2005, 74.1% for year 2010, 70.98% for year 2015, and 78.27% for year 2020. Projections was made for years 2025 and 2030 and it shows a 27.41% decline from the base year by 2025, and a 29.90% decline by 2030. Keywords: Cellular Automata, Markov Chain, Simulation, Supervised Classification

本研究针对奥科穆国家公园（Okomu National Park）2000年至2020年这二十年的时空土地利用变化开展监测与建模，旨在对近期森林覆盖变化进行预测。本研究构建了概率元胞自动机（Probabilistic Cellular Automata, CA）模型，用于模拟土地利用变化并实现预测目标。研究采用2000、2005、2010、2015及2020年的Landsat 7 ETM+卫星影像，通过最大似然监督分类算法将影像划分为森林与非森林两类。各研究年份的分类总体精度分别为98.1838%、97.5169%、96.3325%、91.6647%及94.6124%，对应的总体Kappa系数依次为0.9654、0.9557、0.9524、0.8563及0.9094。基于分类后的影像，本研究计算得到2000-2005、2005-2010、2010-2015及2015-2020四个时段的土地状态转移概率。本研究采用带有冯·诺依曼扩展的摩尔邻域（Moore’s neighborhood with a Von Neumann extension）的概率元胞自动机模型，以2000年作为基准年，对2005、2010、2015及2020年的土地利用变化进行模拟。模拟精度结果显示，2005年模拟精度为77.46%，2010年为74.1%，2015年为70.98%，2020年为78.27%。本研究对2025年与2030年的森林覆盖情况进行了预测，结果表明，相较于基准年，2025年森林覆盖将下降27.41%，2030年将下降29.90%。关键词：元胞自动机、马尔可夫链、模拟、监督分类