Factors affecting the quality of DTM generation in forested areas

Abstract

Airborne laser scanning (ALS) has become an established tool for acquiring digital terrain models (DTM) in forested areas. Even though, there have been several empirical studies on DTM quality with laser scanning, a few studies have focused on factors affecting the quality of DTM generation. This paper analyses especially the effects of the date, flight altitude, pulse mode, terrain slope, forest cover and plot variation on the DTM accuracy at boreal forest zone. The boreal test site was collected with Toposys I and Toposys II in 1998, 2000 and 2003. Since the measurements were recorded at various time of the season, i.e. May 14th 2003 (leaf-off), June 14th 2000 (leaf-on, low development of undergrowth), and September 2nd 1998 (leaf-on, high undergrowth), it was possible to estimate the effect of leaves and undergrowth. In 2003 the flight altitudes of 400, 800 and 1500 m above ground level were used providing nominal pulse densities of 8-10, 4-5 and 2-3 pulses per m2. At boreal forest zone, the random errors of less than 20 cm were obtained in most conditions for non-steep terrain. The increase of flight altitude 400 to 1500 m increased the random error of DTM derivation by 50%. The difference of using first or last pulse caused a similar random error difference. There were systematic shifts in the elevation models derived at various flight altitudes. It is expected that the beam size and sensitivity of the laser system determine this systematic behaviour. Additionally, the systematic shifts between last and first pulse were significant. The difference of DTMs derived at optimum and non-optimal season conditions were typically less than 5 cm for high-density data. In stands consisting of deciduous trees, the seasonal effects were the highest. The random error increased with increasing terrain slope. The effect of forest cover was higher when moving closer to the trunk. The results were site dependent, i.e. the obtained accuracy varied strongly as a function of site conditions.