Mapping Urban Forest Leaf Area Index Using Lidar: A Comparison of Gap Fraction Inversion and Allometric Methods

Abstract:

In urban areas leaf area index (LAI) is a key ecosystem structural attribute with implications for energy and water balance, gas exchange, and anthropogenic energy use. Typically, citywide LAI estimates are extrapolated from those made on forest inventory sample plots through intensive crown measurement and allometric scaling. This is a time- and labor-intensive process yielding coarse spatial resolution results. In this study we generate spatially explicit estimates of LAI using high-point density airborne lidar throughout our study area in downtown Santa Barbara, CA. We implement two theoretically distinct modeling approaches. First, based on hemispherical photography at our 71 field plots, we estimate effective LAI using scan-angle corrected lidar laser penetration metrics (LPM). For our second approach, we adapt existing allometric equations for use with a suite of crown structural metrics (e.g., tree height, crown base height) measured with lidar. This approach allows for estimates of LAI to be made at the individual tree crown scale (ITC). This is important for evaluating fine-scale interactions between canopy and urban surfaces.

The LPM method resulted in good agreement with field estimates (r² = 0.80) and a slope of near unity (β = 0.998) using a model that assumed a spherical leaf angle distribution. Within ITC segments that were automatically delineated using watershed segmentation, lidar estimates of crown structure closely paralleled field measurements (r²=0.87 for crown length). LAI estimates based on the lidar structural variables corresponded well with estimates from field measurements (r² = 0.84). Agreement between the LPM and allometric lidar methods was also strong across the 71 validation plots (r² = 0.88) and among 450 sample points (r² = 0.72) randomly distributed throughout the citywide maps. This is notably higher than the agreement between the hemiphoto and allometric ground-based estimates (r² = 0.56). The allometric approach generally yielded higher estimates except for under closed, broadleaf canopy. The first-order agreement between these two disparate methods may indicate that the error bounds for mapping LAI in cities are small enough to pursue large scale spatially explicit estimation.