Agriculture, Ecosystems and Environment 101 (2004) 155–169 Agricultural landscape change and stability in northeast Thailand: historical patch-level analysis Kelley A. Crews-Meyer ∗ Department of Geography, University of Texas, 210 W. 24th #334, Mailcode A3100, Austin, TX 78712-1098, USA Abstract The temporal nature of agricultural landscape change, where both intra- and inter-annual processes and changes are often at work, renders traditional methods of landscape change assessment not completely effective. Additionally, seasonal and longer-term shifting patterns of cultivation can sometimes appear as permanent landscape change when in fact they actually are simply a local change in spatial arrangement. To address these complexities, this work tests an approach that is longitudinal in character and based upon assessing the structure of landscape change as well as the landuse/landcover (LULC) change. Set in rural northeast Thailand, patch dynamics are examined through use of LULC change trajectories built from an image time series and temporal patterns built from pattern metrics. The given unit of observation is the pixel, and its “life history” is constituted by the values derived from the images of a satellite time series, which are then reconstituted at the patch level for better ecological interpretation. The hypothesis that underpins this approach is that the nature of the trajectory is associated with the function of the land in that patch and in the neighborhood of surrounding patches. Hence, different trajectories of LULC spatial arrangement may suggest, for example, differences in the stability or dynamics of LULC over time and space, which are further suggestive of land sustainability or resilience, or conversely land conversion or dynamism. The study area for this research is a marginalized, agrarian environment in northeast Thailand, a region that has undergone deforestation of upland forests for the cultivation of commercial field crops, intensification of lowland rice for subsistence as well as local and regional sales and global export, and LULC scenarios altering the savanna landscape that serves as the background matrix. The analysis here characterizes the relative stability and temporal dynamics of LULC at the patch level. Pattern metrics calculated at the patch level are assessed as the spatial organization of landscape units that represent: (1) transitional areas of LULC dynamics occurring as peripheral expansion, (2) LULC change from forest to agriculture through deforestation, or (3) agriculture to forest through secondary plant succession, with savanna serving as a transitional matrix. In short, this paper proposes and tests a method for assessing the temporal persistence of LULC through pattern metrics. The method contributes a technique for analyzing the landscape ecology of sites as a function of their stability/dynamics within a scale-explicit context, and contributes to the growing body of work on relating scale, pattern, and process. © 2003 Elsevier B.V. All rights reserved. Keywords: Panel analysis; Change trajectories; Image time series; Pattern metrics; Patch dynamics; Landuse/landcover change; Landscape stability or dynamics ∗ Tel.: +1-512-471-5116; fax: +1-512-471-5049. E-mail address: kacm@uts.cc.utexas.edu (K.A. Crews-Meyer). 1. Introduction Landscape analysis has, in the last century, relied heavily upon remotely sensed data (Lee, 1922; Jensen, 2000). In particular, landuse/landcover (LULC) in- formation derived from a variety of sensor systems 0167-8809/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.agee.2003.09.024