Ecological condition indicators for dry forest: Forest structure variables estimation with NDVI texture metrics and SAR variables

Abstract

The ecological condition of forest ecosystems is degraded. Limited prior research in vegetationhas focused on monitoring ecological condition levels in dry forest at fine scale. We proposeda novel approach to obtain accurate indicators of the ecological condition of the ChacoSerrano forest (Córdoba, Argentina) by estimating forest structure variables (canopy cover (𝐶𝐶),diameter breast height (𝐷𝐵𝐻_𝑠𝑢𝑚), number of woody individuals (𝑁𝑊 ) and two first axesof a principal component analysis (𝑃𝐶1 and 𝑃𝐶2)) as a measure of forest degradation. Toachieve this, first the correlation with two complementary groups of remote sensing deriveddata (texture metrics over Normalised difference vegetation index and SAR-derived data) wasexplored. Then, General linear models (GLM) were constructed using the most correlatedremote sensing derived variables with forest structure variables as predictor variables. Thebest estimation was obtained to 𝐶𝐶 (𝑟2=0.58, rmse=14,5%), followed by 𝐷𝐵𝐻𝑠𝑢𝑚 (𝑟2=0.37,rmse=156.6) and 𝑁𝑊 (𝑟2=0.22, rmse=14.6), with an spatial arrangement consistent with fieldobservations. Moreover, 𝐶𝐶 estimation was more accurate than those at regional and globalscale, and highlights the importance of developing local models in areas that exhibit highecological, geological, and human heterogeneity. In addition, other forest variables could also beevaluated, like floristic composition or others associated with functioning. Results offer valuableinsights for developing management strategies suitable for each condition, and for future studiesregarding the relationship of the mentioned condition and associated natural and anthropicfactors.