**3.3 Summary**

14 Remote Sensing of Biomass – Principles and Applications

Fig. 6. An example of tree crown delineation with airborne lidar data using object oriented

Results show that lidar can detect most of the crowns of dominant and co-dominant trees in mature stands dominated by coniferous species, but finds difficulties for suppressed or subdominant trees, young stands, groups of trees growing in close proximity and for deciduous species. This will have an effect on biomass estimation when aggregating individual tree

Small individuals will contribute relatively less to the overall stand volume than larger trees. However in many situations they may form a significant contribution to total biomass which should be accounted for. Authors such Maltamo *et al.*, 2004 have shown it to be possible to

The detection and estimation of individual tree heights for deciduous species is more difficult because of their more spherical crown shape compared to conifers (meaning that the tree top and crown boundaries can be located with less certainty using a CHM) and the higher probability for the occurrence of more than one apex in each tree. Moreover, crown delineation in deciduous stands becomes more difficult because the crowns of neighbouring

The single tree approach allows individual tree counts, crown volume calculations, canopy closure or single tree height estimates. These are important inputs in order to derive estimates of diameter at breast height (DBH) distributions, volume or biomass. Published studies show that good results can be achieved with 65% to 90% of correct tree counts within conifer stands while for broadleaf trees the results are less accurate and an

Using canopy delineation from airborne lidar data, parameters such as biomass can be determined and spatially located in relation to their surroundings. This also allows the inhibiting effects on biomass accumulation introduced by larger trees over more

predict the height of small trees not detected by lidar using Weibull distributions.

trees often overlap (this is a common situation in conifer stands too).

underestimation of volume is often reported.

disadvantaged neighbours to be observed.

analysis in Definiens Developer (Source: Suárez, 2010).

estimates to a stand level.

In operational terms, current usage of airborne lidar data is generally confined to statistical analysis at stand/forest scales (Næsset, 1997a, b). Although the method has undergone minor adaptations by several authors, this approach has become a standard practice for large-scale inventories in some European countries (Finnish\_Forest\_Association, 2007; Hollaus *et al.*, 2009; Næsset, 2004).

Airborne lidar point clouds can be understood and interpreted intuitively and vegetation can be placed in the context of terrain, access routes and neighbouring competitors. The ability to remove the vegetated surface to reveal the terrain beneath provides valuable information for management purposes and topographic assessment.

Regression equations for biomass estimation may be site specific, however good relationships have been demonstrated using broad class distinctions such as broadleaves, conifers and mixed stands which may be more readily available using optical data or land cover type maps. This enables the spatial distribution of biophysical parameters to be represented over a forest scale which would be impossible using traditional field measurements.

The cost of airborne coverage may prohibit repeat lidar campaigns and so at best, such detailed information is likely to be available infrequently to monitor growth. However, looking to the future, several European countries have undertaken campaigns for partial or complete nation-wide coverage of airborne lidar data including The Netherlands (Duong *et al.*, 2009), Norway, Austria, Switzerland (Swisstopo, 2011) and Finland (Finnish\_Forest\_Association, 2007). A similar commitment of lidar acquisition is in progress in Spain on a region-by-region basis where the value-added benefit of such a resource has been recognised for economic and social invigoration (Dielmo, 2011).

Whilst generally of mid-low point density by necessity of cost and not necessarily intended specifically for vegetation applications, nevertheless, such previously-unobtainable national vertical profile datasets offer potential applications for a multitude of applications including mapping and reducing uncertainty of biomass distribution. Furthermore, as outlined above, airborne lidar data are already playing a significant role in large-scale forest inventory efforts.
