Remote Sensing of Canopy Leaf Area Index in Non-Forested Wetlands

Abstract:

Canopy leaf area index (LAI; one-sided leaf area per unit ground area) is a key instrumental variable used in models of plant-atmosphere carbon and water exchange, greenhouse gas and energy budgets and canopy-based habitats. Multiple studies have measured LAI in upland terrestrial landscapes and explored methods to up-scale field values to regional extents using remote sensing. However, in wetland ecosystems globally, much uncertainty still exists on magnitude of LAI, its spatial and temporal variation and on robust approaches to measure this index in the field and from remote sensing. We assessed LAI in different wetlands of the Sacramento-San Joaquin Delta, California, USA (the Delta) in growing seasons of 2013-2014 and tested its empirical relationships with spectral indices of vegetation function derived from Landsat satellite images. Peak-season site-average LAI ranged from 3.3m²m^-2 in a diked marsh to 6.5m²m^-2 in a young engineered wetland. Results also indicate high within-site dispersion of LAI (coefficient of variation from 0.13 in rice paddy to >0.5 in tall-canopy reed-dominated marshes) attributed to complex surface composition, variable canopy height and non-uniform contribution of litter. Optically measured field LAI significantly correlated (p<0.001) with several Landsat-based indicators of vegetation greenness; however, the strongest univariate relationships explained only 45-50% of LAI variance due to variable canopy characteristics and sub-pixel wetland complexity. Goodness of fit in these relationships improved following corrections based on subpixel spectral unmixing of the green cover fraction. Results indicate that single site-level “mean” LAI values may not sufficiently characterize complex Delta wetland canopies, and models of wetland ecosystem function and greenhouse gas fluxes should incorporate within-site spatial variation in canopy properties. Landsat satellite imagery is promising for regional-scale modeling of LAI, however, simple univariate relationships with spectral indices may need to be extended to multivariate spatial models with sub-pixel corrections for site heterogeneity and phenological characteristics of vegetation.