Estimating landscape fragmentation indices from satellite images: the effect of sensor spatial resolution

Abstract

Fragmentation indices derived from remotely sensed data are being increasingly used for landscape condition assessment and land cover change characterization. However, it is not yet fully understood how fragmentation indices are affected by spatial resolution, and the lack of comparability across scales seriously limits the potential usefulness of this kind of quantitative analysis of landscape patterns. We here consider a wide set of commonly used fragmentation indices and analyze the ability of aggregation filters for replicating the pattern and indices values corresponding to coarser spatial resolution sensors. We analyze simultaneously gathered Landsat-TM and IRS-WiFS satellite images, as well as TM patterns aggregated to coarser resolutions through standard mean and majority filters and through filters that incorporate the specific point spread function of the WiFS sensor. All the images were classified in forested areas, agricultural lands and water bodies for the computation of the fragmentation indices. We show that mean and majority filters tend to produce clearly more fragmented patterns than actual sensor ones. We found that incorporating point spread function in the aggregation process allowed to considerably improve the comparability of fragmentation estimations across spatial resolutions. The biggest improvement was found for indices like number of patches, edge length and mean patch size, which are the most sensitive to changes in spatial resolution and minimum mapping unit. On the contrary, indices like largest patch index, patch cohesion or landscape division were little affected by spatial resolution and did not show significant differences between the aggregation filters considered. Higher aggregation errors were found for water bodies than for forested areas or agricultural lands.Fragmentation indices derived from remotely sensed data are being increasingly used for landscape condition assessment and land cover change characterization. However, it is not yet fully understood how fragmentation indices are affected by spatial resolution, and the lack of comparability across scales seriously limits the potential usefulness of this kind of quantitative analysis of landscape patterns. We here consider a wide set of commonly used fragmentation indices and analyze the ability of aggregation filters for replicating the pattern and indices values corresponding to coarser spatial resolution sensors. We analyze simultaneously gathered Landsat-TM and IRS-WiFS satellite images, as well as TM patterns aggregated to coarser resolutions through standard mean and majority filters and through filters that incorporate the specific point spread function of the WiFS sensor. All the images were classified in forested areas, agricultural lands and water bodies for the computation of the fragmentation indices. We show that mean and majority filters tend to produce clearly more fragmented patterns than actual sensor ones. We found that incorporating point spread function in the aggregation process allowed to considerably improve the comparability of fragmentation estimations across spatial resolutions. The biggest improvement was found for indices like number of patches, edge length and mean patch size, which are the most sensitive to changes in spatial resolution and minimum mapping unit. On the contrary, indices like largest patch index, patch cohesion or landscape division were little affected by spatial resolution and did not show significant differences between the aggregation filters considered. Higher aggregation errors were found for water bodies than for forested areas or agricultural lands.