Introduction Landsliding contributes to the large-scale dynamics of tropical mountains, yet compared to deforestation and fire it remain largely unstudied. One reason for this has been the limited availability of cloud-free imagery capturing meaningful variables that could be linked to biogeochemical cycles and succession at resolutions commensurate with the observed variability of landslide sizes. The acquisition of vegetation height data by the Global Ecosystem Dynamic Investigation GEDI in combination with Landsat Analysis Ready Data (ARD) may offer the possibility to examine the integrated response of landslides on ecosystems at landscape scales. Objectives Focusing on the Sierra de Las Minas of Guatemala (SLM) we address two sets of objectives. First, develop a regional model linking metrics derived from Landsat data with GEDI’s tree height to produce a temporal series of continuous, high-resolution tree height maps. Second, characterize patterns of vegetation re-growth and biomass accumulation in areas affected by landslides. Methods We combined three approaches to address the two objectives. First, we built a landslide geodatabase using historical imagery (mid 1960s – 2020) that was orthorectified, mosaicked, and classified using ERDAS’ Imagine Feature Extraction tools. Second, we used Boosted Tree (BRT) Regressions to model the relationship between the Pheno metrics derived from GLAD’s Landsat ARD and GEDI’s v2 RH95. This model was used to predict RH95 over the SLM for each year included in the Landsat series (2000-2020). Lastly, we examined the temporal variation of tree height in our mapped landslides to understand changes in biomass and forest structure during succession. Results A first Boosted Tree (BRT) Regression predicted tree height over our study area. Although there was a good agreement between the predictions and the RH95 heights, they either sub-estimated the height of tall trees or over-estimated the height of short vegetation. In an area like the SLM where trees may reach up to 50 m, the sub-estimation of tree height may have important consequences for the estimation of biomass and characterization of forest structure. Similarly, the over-estimation of small-stature vegetation developing in disturbed areas or characteristic in certain areas may result in vegetation re-growth or biomass accumulation rates that are not indicative of existing conditions. Implications A full evaluation of the temporal predictions may help determine the utility of this data to examine the integrated response of landslides on ecosystems at landscape scales.
GEDI, Landsat, landslides, succession, mountains, remote sensing