Is a Canopy Height Model enough for estimating forest biomass with Airborne Laser Scanning?

The data retrieved by Airborne Laser Scanning demonstrated to be a valuable source of information for estimating forest variables such as biomass or growing stock volumes. Model-based estimations with parametric or non-parametric methods are used for calculating a large number of metrics on the basis of laser point returns which fall inside the sampling plots carried out on the ground. A model is developed on the basis of ALS returns to predict the forest variable and then is applied to all forest pixels in a given study area. This approach is known as echoes-based since it requires the availability of raw ALS returns and for this reason is time consuming and computational intensive for the need of manipulating large amount of ALS returns. A simplified methodology is the so called area-based approach. Following such approach ALS data are elaborated only for producing a standard raster Canopy High Model (CHM). Simplified metrics are calculated for field plots on the basis of the CHM and then used for setting up a model that is then applied to forest pixels, exactly as it happens in the echoes-based approach. This simplified methodology is less time consuming and less computational intensive since in most of the cases CHM can be easily calculated with a simple map algebra operation between a Digital Surface Model and a Digital Elevation Model, both in raster format, which are usually provided by the subject in charge of the ALS acquisition. Thus the area-based approach can be applied without manipulating the ALS returns and is the only possibility when raw ALS echoes are not available. In the last years an interesting debate emerged in the scientific community about pros and cons of the two methods. The purpose of this paper is aimed at comparing the spatial explicit estimation of forest total biomass with ALS data contrasting echoes-based and area-based approaches. The test area of 36380 ha, located in Regione Molise (Central Italy), was covered by a network of hexagons of 1 km2 each. In the first phase of sampling, a point was randomly selected in each hexagon in accordance with the TSS scheme. The first-phase points were classified as “forest” and “non forest” on the basis of an aerial high resolution photography acquired in 2006. In the second phase, a sample Q of points was selected from the first phase 197 forest points by means of SRSWOR. A field campaign was carried out in the period 2009-2011. Between the 62 points of Q. For each point a circular plot of 13 m radius was created around the point and the total biomass was measured (herbs, bushes and trees). In June 2010 the study area was covered by an ALS survey carried out by an Optech Gemini LiDAR. The maximum scanning angle was 15°, the frequency of pulses was 70 KHz resulting in an average density of 1.5 returning echoes per m2. We implemented the echoes-based approach calculating a total of 39 metrics on the 62 field plots and the area-based approach on the basis of a CHM with a resolution of 1 m. Both a parametric (by linear regression) and a non-parametric (by k-nearest neighbors) approaches were applied for a spatial explicit estimation of forest biomass with output pixels of 23 x 23 meters. This contribution presents the results are contrasting the estimation performances of echoes-based vs. area-based approaches compared to design-based estimation as a common standard reference.