ALBERT

Description: In northern Argentina, the assessment of degraded forests is a big challenge for both science and practice, due to their heterogeneous structure. However, new technologies could contribute to mapping post-disturbance canopy cover and basal area in detail. Therefore, this research assesses whether or not the inclusion of partial cover unmanned aerial vehicle imagery could reduce the classification error of a SPOT6 image used in an area-based inventory. BA was calculated from 77 ground inventory plots over 3944 ha of a forest affected by mixed-severity fires in the Argentinian Yungas. In total, 74% of the area was covered with UAV flights, and canopy height models were calculated to estimate partial canopy cover at three tree height classes. Basal area and partial canopy cover were used to formulate the adjusted canopy cover index, and it was calculated for 70 ground plots and an additional 20 image plots. Four classes of fire severity were created based on basal area and adjusted canopy cover index, and were used to run two supervised classifications over a segmented (algorithm multiresolution) wall-to-wall SPOT6 image. The comparison of the Cohan’s Kappa coefficient of both classifications shows that they are not significantly different (p-value: 0.43). However, the approach based on the adjusted canopy cover index achieved more homogeneous strata (Welch t-test with 95% of confidence). Additionally, UAV-derived canopy height model estimates of tree height were compared with field measurements of 71 alive trees. The canopy height models underestimated tree height with an RMSE ranging from 2.8 to 8.3 m. The best accuracy of the canopy height model was achieved using a larger pixel size (10 m), and for lower stocked plots due to high fire severity.