Projects
Grad Corner
An application for graduate students to emotionally support each other.
Visualization of Stochastic Harvest Planning
Assessing Biodiversity indicator using UAV remote sensing
This project explores explored the capability of a fixed-wing UAV, namely the eBee drone, imagery to map forest canopy gaps and derive some forest parameters such as biodiversity indices, habitat trees, basal area, canopy height, deadwood, etc. using handy techniques in a test area of 240 ha of natural reserve of Lago di Vico in Central Italy. The mapping revealed that forest shaded canopy gaps can be faithfully extracted from UAV RGB color images. Estimation of forest features using canopy gaps as a proxy led to disparate results. Best results were obtained for understorey data with R2 going up to 0.87 and intermediate results were observed in living trees data with R2 of over 0.74. The approach failed to estimate the deadwood. Additionally, from the three forest types available in the study area, best results were observed in mixed forest while Beech forest had the poorest ones and Oak forest displayed intermediate results. UAV true color remote sensing presents a high potential for forest inventory, forest monitoring, and biodiversity assessment.
Efficient soil depth sampling scheme using geostatistics: Case of Mamora forest in Morocco
Soil is the most important factor precluding cork oak regeneration in the Maamora forest. Regeneration and reforestation success rate depends on upper soil layer thickness, which is sandy and resting above a clay layer. This study was conducted in order to assess sand thickness and to recommend a reasonable and efficient sampling grid that will minimize mapping cost. To achieve this goal, 1912 plots have been collected according to a systematic sampling grid of 500 m. In addition, and to better understand sand thickness variability, another sampling scheme concerned one stand. It consists of systematic points located at 100 m distance. The data have been processed using ordinary kriging. The results show that sand thickness is variable and is not related to terrain slope. Also, within a sampling grid of 500 m, only around 40% of the sand thickness is spatially auto-correlated. Furthermore, there is an anisotropy in the variation of the thickness of the sand. As conclusion, a systematic sampling grid with plots within 200 to 350 m distance seems to be well suited