Publications

Mondal, S; Jeganathan, C (2022). Effect of scale, landscape heterogeneity and terrain complexity on agriculture mapping accuracy from time-series NDVI in the Western-Himalaya region. LANDSCAPE ECOLOGY, 37(11), 2757-2781.

Abstract
Context Accurate mapping of mountain agriculture is important for both a food-security and an environmental degradation perspective. Remote sensing provides a large and diverse data source for these complex regions and enables large-scale mapping of mountain agriculture. However, obtaining good mapping accuracy is challenging in mountain terrain. Objectives The main objective is to analyse the effect of mapping scale or resolution, landscape heterogeneity and terrain complexity on the accuracy of agriculture extraction in the Western-Himalayan region. Methods A similarity-based technique was applied on time-series MODIS NDVI for extracting agriculture area at 250 m, 500 m and 1000 m resolution. Further, Pareto boundary-based accuracy was estimated from reference high-resolution agriculture data. The agriculture area extracted at multiple scales was utilized to compute landscape heterogeneity indicators, and ASTER DEM was used to estimate the terrain complexity indicators. Finally, the dependency of mapping accuracy on landscape heterogeneity and terrain complexity were analyzed through linear regression. Results Mapping accuracy gradually decreases with coarseness of scale mainly due to higher omission error. Spatial visualization of landscape heterogeneity and mapping accuracy reveals that higher patch fragmentation increases omission error causing lower accuracy. Regression analysis revealed that heterogeneity has a negative relationship with mapping accuracy such that higher fragmentation lowers the accuracy. Terrain complexity significantly (p < 0.05) influences the mapping accuracy only at 250 m resolution. Conclusion Analysis explains that the effect of scale on mapping accuracy for mountain agriculture mapping is mainly controlled by landscape heterogeneity such that patch fragmentation is the cause for introducing negative bias towards mapping accuracy with the coarseness of resolution.

DOI:
10.1007/s10980-022-01533-6

ISSN:
1572-9761