Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Allemagne → 2019 → Assessment of degraded forests supported with UAV imagery towards planning rehabilitation strategies : case study in the Argentinian Yungas

Albert-Ludwigs-Universität Freiburg im Breisgau (2019)

Assessment of degraded forests supported with UAV imagery towards planning rehabilitation strategies : case study in the Argentinian Yungas

Rossi, Fernando C.

Titre : Assessment of degraded forests supported with UAV imagery towards planning rehabilitation strategies : case study in the Argentinian Yungas

Auteur : Rossi, Fernando C.

Université de soutenance : Albert-Ludwigs-Universität Freiburg im Breisgau

Grade : Doctor rer. nat. 2019

To assess and rehabilitate degraded forests are global challenges. Due to the various causes and severity of degradation, they typically show a very diverse structure, which makes acquiring a detailed inventory, as a first step toward rehabilitation measures, difficult and costly. Area-based inventories based on satellite imagery are a well-established methodology to assess and classify the forest cover, but the information obtained is often not detailed enough to fulfil the needs of site-adapted rehabilitation in degraded forests with a highly di-verse structure. Furthermore, due to the great variability of the forest structure, a great number of ground sample plots are necessary to establish statistically sound predictions of structural parameters. New technologies, such as unmanned aerial vehicles (UAV), allow the acquisi-tion of detailed images, but until now they are mostly used to estimate forest attributes only in small scale applications, and if a precise digital terrain model (DTM) is available, since optical sensors cannot sense terrain under dense can-opy covers. However, the frequent canopy gaps on degraded forests would allow the partial assessment of the terrain, and the combination of wall-to-wall satellite imagery with partial cover UAV imagery would allow larger scale in-ventories of acceptable accuracy and reliability without the need for an in-creased number of ground sample plots. This approach was developed and tested in this research, and applied to a 3944 ha case study in the Argentinian Yungas Cloud Forest (named Florestoona), which was degraded by a mixed-severity fire in 2013. Therefore, as a first objective, this research assesses whether or not the inclu-sion of partial cover unmanned aerial vehicle imagery could reduce the classi-fication error of a SPOT6 image used in an area-based inventory of a forest af-fected by a mixed-severity fire. Basal area (BA) was calculated from ground inventory and was used to run a first classification of the satellite image (BA-based classification). Then, BA was correlated to partial canopy cover and dif-ferent tree heights, calculated from AUV imagery derived canopy height mod-els (CHMs), in order to formulate the adjusted canopy cover index (ACCI) and run a second image classification (ACCI-based classification). The ACCI-based classification did not improve the classification error in comparison with the BA-based classification, but it achieved more homogeneous strata, since the inclusion of forest structure parameters derived from AUV imagery allow to correctly classify stands which have similar reflectance but are very different in regard to their structure and stock. Additionally, two resolution DTMs were used to calculate CHMs, which were assessed against measurements in the field at different degrees of forest degradation. CHMs underestimated tree height with a root mean squared error (RMSE) ranging from 2.8 to 8.3 m, having the most accurate estimation with lower resolution DTMs (10 m/pixel) and more degraded forests. Once a detailed map of the degraded forest was available (stands of 3.129 m2 on average were delineated), it was the second objective of this research to es-tablish meaningful forest management units (FMU) in order to plan and im-plement forest rehabilitation measures. FMUs should be big enough to facili-tate operations, but also homogenous on forest structure and stock so that uni-form adapted rehabilitation measures could be applied. The aim was to devel-op and demonstrate a method to establish meaningful FMUs based on the spa-tial distribution of ACCI values as a metric of forest structure and stock. There-fore, close and similar features were clustered through the use of the tool Hot Spot Analysis (Getis-Ord Gi*) from the environment Arc-GIS. Clustering on a multi-scale analysis was conducted to test spatial autocorrelation on neigh-bouring areas of 10, 20, 30 and 40 ha, resulting in threshold radii of 178, 252, 309 and 357 meters, respectively. The tool demonstrated to significantly aggregate between 30.7 % (at 178 m) to 60.8 % (at 357 m) of the area into either cold or hotspots, which would be the basis to delineate FMUs, where adaptive rehabil-itation measures can be planned. In the remaining areas, the structural parame-ters were randomly distributed, thus a conventional approach to identify FMUs by expert knowledge should be applied. The results reported a trade-off be-tween the gain in area of the FMUs and the loss of their homogeneity. The in-correctly clustered area increases from 7.1 % to 19.1 %, from 178 to 357 m dis-tance thresholds. Furthermore, in this dissertation, adaptive rehabilitation strategies for each type of FMU (with concentration of high, low, and randomly distributed ACCI values) are analysed and discussed based on the result obtained from a ground inventory carried out on enrichment planting (EP) experiments established since 2004 in Florestoona. The experimental plots were planted without specific statistical design, with the species Toona ciliate M. Roem., Melia azedarach L., and Tipuana tipu Kuntze. The results reported that the species would need a rotation of 15, 10, and 20 years respectively, in order to reach 40 cm DBH, which is the diameter of harvest. However, those results overestimate the performance of the EP since due to the line or strip design of the EP, up to 80 % of the EP areas are considered as border areas where, due to edge effects, growth is lower than in the core, where the measurements took place. Finally, for a partial area of Florestoona, detailed recommendations for imple-menting adapted rehabilitation measures in selected FMUs are given as an ex-ample based on the findings of this research, and on an extensive literature re-view about the origins and actual implementation of EP in the region. It shows how the concepts and methods developed and applied in this research can support the practical implementation of adapted rehabilitation measures.


Version intégrale (6,54 Mb)

Page publiée le 28 février 2020