Conversion of land to agriculture has transformed vast landscapes and altered fundamental ecosystem processes. Ecology in the Anthropocene is driven by such changes, along with climate change. Technological advances in remote sensing have made monitoring landscape changes more feasible, but the volume of data and interpretation create computational challenges, particularly when very high resolution data are necessary. This is likely for small, nonflying organisms such as amphibians who interact with the landscape at a fine scale. Amphibians are notoriously sensitive to their abiotic environments, and persistence depends on availability of breeding sites, habitat in the surrounding landscape, and access to sites that provide refugia during droughts. Demographic connectivity among portions of a landscape requires continuous habitat suitable for dispersal.
We used remote sensing data to quantify changes in the landscape, particularly in features including grassland/pasture, riparian stretches, tree rows, and wetlands, that facilitate dispersal. We used National Land Cover Data (NLCD) and annual aerial photography from the USDA National Agricultural Imagery Program (NAIP). NAIP imagery is much higher resolution (0.6 - 1m vs. 30m for NLCD), but must be classified to analyze landscape composition, which we accomplished using object-based image analysis (OBIA). We compared accuracy of classified NAIP with NLCD from the same years and estimated potential connectivity based on simple assumptions about habitat suitability for movement.
Results/Conclusions
NLCD is convenient because it is already classified but is coarser resolution and the classification accuracy may not be sufficient to reliably detect important fine features. On our landscapes, accuracies of NLCD ranged from 70-75%. OBIA of NAIP images classified the landscape more accurately (88.4-93.2%) than two pixel-based classification methods (70.9-77.3%) and was also more accurate than NLCD. Notably, smaller features visible in NAIP imagery were not detected by NLCD, including known breeding sites of amphibians, riparian areas, and tree rows. Both data sources revealed changing land cover composition from 2009-2018 consistent with USDA statistics, with declining grassland and increasing cropland, but with different proportions and the opposite trend in potential connectivity. NLCD overestimated wetlands and crops and underestimated grassland. Wetland cover increased slightly, but wetlands on NAIP maps appeared to become functionally less connected for amphibians, based on assumed relationships between land cover and constraints on movement. Changes in landscape composition estimated from classified NAIP imagery predict increasing resistance to movement for animals such as amphibians that have limited ability to transit crop fields.