Vehicle-mounted laser scanning – a viable alternative to field surveys?

Access to remote sensing data is growing rapidly, but for it to be useful in new forest decision-support tools, high-quality reference data is required. So, how accurate are new digital methods for collecting reference data compared with traditional field inventory? A study shows that vehicle-mounted laser scanning can be a viable alternative.

A car equipped with a mobile laser scanning system is driving along winding gravel roads in a forested area outside Skara. As the car drives, it scans the trees in the forest in order to collect reference data for a research project that trains models to estimate the characteristics of individual trees.

This data collection is part of a PhD project within Mistra Digital Forest, which has been investigating how effective this type of reference data can be for estimation of tree trunk volume and diameter at breast height. The focus of the project has been to compare two models that use remote sensing data in order to make these estimates. One model used reference data from vehicle-mounted laser scanning, and the other used reference data from traditional field inventory.

- The reference data we obtain using these different collection methods yielded equivalent results in terms of both the accuracy of the estimates, and the variables on which the models are trained. Vehicle-mounted laser scanning is an interesting alternative for forestry companies looking to save time and money, says Raul de Paula Pires, a PhD student in forest resource management at SLU and the researcher who led the study.

Vehicle-mounted laser scanning can replace field surveys – but both methods lead to systematic errors 

The laser scanner mounted on the vehicle collected data from trees at distances of between 20 and 40 metres from the roadside. Equivalent field data was collected from 265 sample plots, systematically distributed across the same area in Remningstorp, outside Skara. One key difference was that the vehicle-mounted method was able to collect data from significantly more trees – over 6,000 compared to around 2,000 for the field inventory.

– We can see that more data does not automatically mean better results; it is most important that the data is representative of the forest. The results show that vehicle-mounted laser scanning can be a viable alternative to traditional field inventory, but that both methods lead to systematic errors. Spruce, which accounted for over 85 per cent of the trees in the area, was estimated more accurately than pine and deciduous trees, and small trees were often not detected at all. Here, there is a risk that the total timber volume is underestimated, says Raul de Paula Pires.

Eva Lindberg, senior lecturer at the Department of Forest Remote Sensing at SLU, outlines where research needs to be focused in the future:

– Vehicle-mounted laser scanning has the potential to become an important source of reference data for estimates derived from high-resolution remote sensing data. However, more research is needed to investigate the effect of vehicle-mounted laser scanning only measuring within a few dozen metres of the roads. Trees often look different there, compared to the rest of the forest. Will the estimates be accurate on average?