4 May 2015 Science Briefs

SNAMP PUB #37: Lidar with multi-temporal MODIS provides a means to upscale predictions of forest biomass

Article Title: Lidar with multi-temporal MODIS provides a means to upscale predictions of forest biomass

Authors: Le Li, Qinghua Guo, Shengli Tao, Maggi Kelly, and Guangcai Xu

Research Highlights:


  • In this study, we use airborne Lidar to “bridge the scale gap” between satellite-based and field-based studies, and evaluated satellite derived indices to estimate regional forest aboveground biomass.

  • MODIS and Lidar data can be used together to estimate forest aboveground biomass at larger scales than with Lidar data alone.

Background:
Forests play a key role in the global carbon cycle, and forest above ground biomass is an important indictor to the carbon storage capacity and the potential carbon pool size of a forest ecosystem. Satellite remote sensing provides abundant observations to monitor forest coverage, but validation of coarse-resolution biomass estimates derived from satellite observations is difficult because field measurements are gathered at very fine-scales, and the footprints of satellite observations are course-scale.

Methods:
We used field data gathered by the Fire and Forest Ecosystem Health SNAMP science team. These plots were placed in a regular grid 500 m apart. Each circular study plot had a radius of 12.62 m. Within each plot, tree species, diameter at breast height (DBH), tree height, and height to live crown base (HTLCB) for trees over 5cm DBH were recorded. These data gathered at the plots were used with regional allometric equations from the USDA Forest Service Forest Inventory and Analysis (FIA) project to estimate biomass for individual trees, and to summarize biomass for each plot. Plot-level biomass estimates were used with lidar and satellite imagery to estimate biomass using correlation models. We evaluated several correlation models, including linear regression. Typically these correlation models are evaluated using two measures: A good model fit has a high correlation coefficient (R2) and low root-mean-squared-error (RMSE). We first compared field data with lidar data to evaluate which lidar metrics would be most useful to estimate biomass. Next, at the regional scale, we compared lidar data with vegetation indices derived from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery. Vegetation indices are essentially images depicting where vegetation is most abundant. Since MODIS imagery is captured frequently, we evaluated how well vegetation indices from a single date would predict biomass, and compared that to vegetation indices that accumulated vegetation values from multiple dates.

Results:
Accurate models that used Lidar data with plot-scale measures of tree height and tree quadratic height were developed. The linear regression model achieved the best performance (R2 = 0.74; RMSE = 183.57 Mg/ha). At the regional scale, multi-temporal MODIS vegetation indices were found to be good predictors for forest aboveground biomass modeling. The multi-temporal vegetation index derived from 1 km resolution MODIS imagery produced the highest model fit: with R2 = 0.74 and RMSE = 13.4 Mg/ha. This is likely due to the fact that the multi-temporal vegetation index captures well the changing forest over a growing season.

Conclusions:


  1. Forest aboveground biomass can be estimated using linear regression models using allometrically-derived, plot-scale measures of biomass and lidar measures.

  2. MODIS and Lidar data can be used together to estimate forest aboveground biomass at larger scales than with Lidar data alone.
    The multi-temporal vegetation index derived from 1 km resolution MODIS imagery was the best input data to estimate biomass at regional scales.

Full Reference:
Li, L., Q. Guo, S. Tao, M. Kelly, and G. Xu. 2015. Lidar with multi-temporal MODIS provide a means to upscale predictions of forest biomass. ISPRS Journal of Photogrammetry and Remote Sensing. 102: 198–208.

The full paper is available here.

For more information about the SNAMP project and the Spatial team, please see:
Spatial Team.

No Comments Yet

Post a New Comment