Grabow Maik, MSc

Direct quality estimation studies of forest roads using remote sensed data are still rare. Research focussing on deriving indirect factors about road quality such as wetness or soil bearing capacities are prone to uncertainty of estimates. Altough processes of direct 3D roadsurface measurements and assesements exist, those systems mostly rely on ground-based collection methods, which are laborious and expensive for big forest road networks. Because of the increased quality of airborne laser scanned data (ALS) over the last decades, it was researched how this development may be used for forest road infrastructure assesement. The present study uses ALS data to estimate the occurrence of road damage in the form of concave depression. Maximum damage depth measurents were collected in a test area near Ilomantsi (Finland) and used to fit a linear mixed effect model to a subset of variables derived from ALS data as fixed factor and location as random factor. Simultaniously, depth measurements were carried out in a digital terrain model and compared to the real depth values to gain insight into maximum reachable accuracies. It was detected that ALS data tend to underestimate road damage depressions, limiting results in follow up models. Our results also indicate that road location could explain up to 30% of the variablity in our models. In the most optimal case, a linear mixed effect model could achieve an R2 of 0.87 with and 0.66 without the random factor having a residual mean standard error (RSME) of 1.9 to 2.7 cm on unvegetated forest roads. The best performing model determined in our study using ALS derived variables reached an R2 of 0.58 and 0.44 with an RSME of 1.8 to 2.4 cm. The work conducted gave insight into the culprits and limits of road depression depth estimates using ALS data. Future research should be conducted to change the scope of the analysis to an evaluation of bigger road segments and to investigate how road location could be utilized to achieve results with better accuracies.

Direct quality estimation studies of forest roads using remote sensed data are still rare. Research focussing on deriving indirect factors about road quality such as wetness or soil bearing capacities are prone to uncertainty of estimates. Altough processes of direct 3D roadsurface measurements and assesements exist, those systems mostly rely on ground-based collection methods, which are laborious and expensive for big forest road networks. Because of the increased quality of airborne laser scanned data (ALS) over the last decades, it was researched how this development may be used for forest road infrastructure assesement. The present study uses ALS data to estimate the occurrence of road damage in the form of concave depression. Maximum damage depth measurents were collected in a test area near Ilomantsi (Finland) and used to fit a linear mixed effect model to a subset of variables derived from ALS data as fixed factor and location as random factor. Simultaniously, depth measurements were carried out in a digital terrain model and compared to the real depth values to gain insight into maximum reachable accuracies. It was detected that ALS data tend to underestimate road damage depressions, limiting results in follow up models. Our results also indicate that road location could explain up to 30% of the variablity in our models. In the most optimal case, a linear mixed effect model could achieve an R2 of 0.87 with and 0.66 without the random factor having a residual mean standard error (RSME) of 1.9 to 2.7 cm on unvegetated forest roads. The best performing model determined in our study using ALS derived variables reached an R2 of 0.58 and 0.44 with an RSME of 1.8 to 2.4 cm. The work conducted gave insight into the culprits and limits of road depression depth estimates using ALS data. Future research should be conducted to change the scope of the analysis to an evaluation of bigger road segments and to investigate how road location could be utilized to achieve results with better accuracies.