*This work was partially supported by NASA Grant No. NAG5-9331 to Dr. Michael W. Binford. We are grateful to Emilio Bruna for some useful discussions that helped to refine the contents of this article and to Dr. Peter Waylen who provided invaluable assistance with the technical difficulties associated with programming the neutral landscape model. We are also grateful to Dr. Kelley Crews-Meyer and Dr. Kenneth Young for organizing a special session on Landscape Ecology and GIS at the Centennial AAG Meeting in 2004 from which this article originated and for their continued and tireless efforts to have the outputs from this meeting published together in a top geography journal.
Linking Spatial and Temporal Variation at Multiple Scales in a Heterogeneous Landscape*
Version of Record online: 20 OCT 2006
The Professional Geographer
Volume 58, Issue 4, pages 406–420, November 2006
How to Cite
Southworth, J., Cumming, G. S., Marsik, M. and Binford, M. W. (2006), Linking Spatial and Temporal Variation at Multiple Scales in a Heterogeneous Landscape. The Professional Geographer, 58: 406–420. doi: 10.1111/j.1467-9272.2006.00578.x
- Issue online: 20 OCT 2006
- Version of Record online: 20 OCT 2006
- Initial submission, March 2005; revised submission, January 2006; final acceptance, April 2006.
Anthropogenic, ecological, and land-surface processes interact in landscapes at multiple spatial and temporal scales to create characteristic patterns. The relationships between temporally and spatially varying processes and patterns are poorly understood because of the lack of spatiotemporal observations of real landscapes over significant stretches of time. We report a new method for observing joint spatiotemporal landscape variation over large areas by analyzing multitemporal Landsat data. We calculate the spatiotemporal variation of the Normalized Difference Vegetation Index (NDVI) in the area covered by one Landsat scene footprint in north central Florida, over spatial windows of 104–108 m2 and time steps of two to sixteen years. The correlations, slopes, and intercepts of spatial versus temporal regressions in the real landscape all differ significantly from results obtained using a null model of a randomized landscape. Spatial variances calculated within windows of 105–107 m2 had the strongest relationships with temporal variances (regressions with both larger and smaller windows had lower coefficients of determination), and the relationships were stronger with longer time steps. Slopes and y-intercepts increased with window size and decreased with increased time step. The spatial and temporal scales at which NDVI signals are most strongly related may be the characteristic scales of the processes that most strongly determine landscape patterns. For example, the important time and space windows correspond with areas and timing of fires and tree plantation harvests. Observations of landscape dynamics will be most effective if conducted at the characteristic scales of the processes, and our approach may provide a tool for determining those scales.