Cellular Automata Modeling of Land-Use/Land-Cover Dynamics: Questioning the Reliability of Data Sources and Classification Methods

Abstract

Based on four time intervals within a forty-year period of observation, we construct land-use/land-cover (LULC) maps and estimate the transition probabilities between six LULC states. The maps and transition probability matrices (TPMs) were built based on the high-resolution aerial photos and 30-m multispectral Landsat images for the same years. We considered the TPM constructed from manual classification of the aerial photos as a reference and compared it to the TPM constructed from the Landsat image classified with several methods: mean-shift segmentation followed by random forest classification and three pixel-based methods popular in cellular automata (CA) studies: K-means, iterative self-organizing data analysis techniques (ISODATA), and maximum likelihood. For each classification method, the TPMs were constructed and compared to the TPMs for the aerial photos. We prove that the goodness-of-fit of maps obtained with the three pixel-based methods was insufficient for estimating the LULC TPM. The LULC maps obtained with the object-based classification fit well to those based on the aerial photos, but the estimates of TPM were yet qualitatively different. This article raises doubts regarding the adequacy of Landsat data and standard classification methods for establishing LULC CA model rules and calls for the careful reexamination of the entire land-use CA framework. We appeal for a new view of the CA modeling methodology: It should be based on a long-term series of carefully validated LULC maps that portray different types of land-use dynamics and land planning systems over long and representative periods of population and economic growth.

我们根据在四十年观察期间的四次间隔, 建构土地使用/地表覆盖 (LULC) 的地图, 并评估六个 LULC 州的转移可能性。这些地图与转移概率矩阵 (TPMs), 是根据同年的高分辨率航摄照片和三十公尺的多光谱地球卫星影像建构之。我们考量从航摄照片的人工分类所建立的TPM作为参照, 并将其与以若干方法进行分类的地球卫星影像所建构的TPM进行比较：均质平移分割, 以及随后的随机森林分类, 还有细胞自动机 (CA) 研究中以像素为基础的的三大流行方法：K-平均数, 迭代自组织数据分析演算法 (ISODATA), 以及最大化可能性。我们对每个分类方法建立 TPMs 并与航摄照片的TPMs进行比较。我们証实, 透过三个以像素为基础的方法所取得的地图适合度, 对评估 LULC TPM 而言并不充足。以物件为基础的分类法所获得的LULC地图, 相当符合根据航摄照片的地图, 但TPM的估计在数值上却仍不同。本文提出有关地球卫星影像和标准分类方法在建立 LULC CA 模型规则上的适切性之质疑, 并呼吁再次对整个土地使用 CA 架构进行细緻的检验。我们恳求对 CA 模式化方法的崭新观点：它应该根据描绘在人口及经济成长的长期代表週期中, 不同种类的土地使用动态和土地使用规划系统、且经过长时间仔细验证的 LULC 地图。

Con base en intervalos temporales de cuatro dentro de un período de observación de cuarenta años, construimos mapas de uso del suelo/cobertura de la tierra (LULC) y calculamos las probabilidades de transición entre seis LULC estatales. Los mapas y las matrices de probabilidades de transición (TPMs) fueron construidos con base en aerofotos de alta resolución e imágenes Landsat multiespectrales de 30-m para los mismos años. Tomamos como referencia la TPM construida a partir de la clasificación manual de las aerofotos y la comparamos con la TPM construida de la imagen Landsat clasificada con varios métodos: segmentación de cambio mediano seguida por clasificación forestal aleatoria, y tres métodos basados en pixeles que son populares en los estudios de modelos celulares autómatas (CA): el K-means, las técnicas de análisis de datos iterativos auto-organizados (ISODATA) y la probabilidad máxima. Para cada método de clasificación, las TPMs fueron construidas y comparadas con las TPMs para las aerofotos. Probamos que la bondad de ajuste de los mapas obtenidos con los tres métodos basados en pixeles era insuficiente para calcular la TPM de LULC. Los mapas de LULC obtenidos con la clasificación basada en objeto encajaron bien con los basados en aerofotos, aunque los cálculos de la TPM eran todavía cualitativamente diferentes. Este artículo levanta dudas sobre la idoneidad de los datos Landsat y los métodos de clasificación corrientes para establecer reglas del modelo CA LULC y clama por un cuidadoso reexamen de todo el marco CA de uso del suelo. Somos partidarios de una nueva visión de la metodología modeladora de los CA: Esta debe basarse en una serie a largo plazo de mapas LULC cuidadosamente validados que representen los diferentes tipos de dinámica de uso del suelo y sistemas de planificación de la tierra sobre largos y representativos períodos de crecimiento de la población y la economía.

Key Words:

• cellular automata
• Landsat images
• land-use/land-cover changes
• Markov transition probabilities matrices
• validation of remote sensing classification methods

关键词：:

• 细胞自动机, 地球卫星图像, 土地使用／地表覆盖变迁, 马尔科夫转移概率矩阵, 遥感分类方法的验证。

Palabras clave:

• celulares autómatas
• imágenes Landsat
• cambios de uso del suelo/cobertura del suelo
• matrices Markov de probabilidades de transición
• validación de métodos de clasificación por percepción remota

Acknowledgments

We are grateful to two anonymous reviewers for helpful comments on earlier drafts of the article. Many thanks to Carole Shoval for language assistance and editing.

There is always a chance that we are wrong in our choice of method's parameters, RS software, or interpretation of results. Understanding the importance of our results for the future of CA modeling, we make our data and results available at https://drive.google.com/open?id=0B_OK-4hDBIH-Y3hObXV5QkFGaTg. The site contains initial data sets, full transition matrices for six land-use types, and other supplementary materials. The reader is cordially invited to employ all possible classification methods and to upload their classification results for comparison.

Additional information

Notes on contributors

Yulia Grinblat

YULIA GRINBLAT is a PhD candidate at the Porter School of Environmental Studies and a member of Geosimulation and Spatial Analysis Laboratory at the Department of Geography and Human Environment, Tel Aviv University, Tel Aviv 69978, Israel. E-mail: [email protected]. Her main interests include spatial–temporal analysis and modeling of land-use changes along fringe landscapes.

Michael Gilichinsky

MICHAEL GILICHINSKY is a remote sensing specialist and at the time of the presented research he was a research fellow at the Samaria and Jordan Rift R&D Center, Ariel University, Ariel 40700, Israel. E-mail: [email protected]. His research interests include the applications of optical remote sensing methods for land use and land cover mapping.

Itzhak Benenson

ITZHAK BENENSON is Professor of Geography and head of the Geosimulation and Spatial Analysis lab in the Department of Geography and Human Environment, Tel Aviv University, Tel Aviv 69978, Israel. E-mail: [email protected]. His research includes study of big urban data, modeling of urban land use and residential dynamics, impact of local and regional plans, use of public transport and parking in the city, vehicle–pedestrian interactions, and road accidents.