Please use this identifier to cite or link to this item:
|Title:||The role of environmental filters in Brazilian savanna vegetation dynamics|
Random forest algorithm
|Citation:||CORDEIRO, N. G. et al. The role of environmental filters in Brazilian savanna vegetation dynamics. Forest Ecology and Management, [S.l.], v. 500, Nov. 2021.|
|Abstract:||Climate, topography and edaphic characteristics are important factors for plant community dynamics when considering spatial and temporal scales. Efforts on vegetation dynamics in tropical regions are mostly focused on forests, with far less attention dedicated to open vegetation ecosystems. This study aimed to understand how demographic rates (based on models and maps) are affected by climate, soil and terrain variables in the Brazilian savanna, the Cerrado, which is a hotspot for the world biodiversity conservation. To do so, we used forest inventory data collected from 354 plots (10 × 100 m) distributed in the savanna of Minas Gerais State, southeast of Brazil, in the years of 2005–2006 and 2010–2011. Next, we calculated the rates of mortality, recruitment, net change in number of individuals, basal area loss and gain, and net change in basal area for the plots. We used the Random Forest (RF) algorithm to model the demographic rates in function of climate, soil and terrain variables obtained from the WorldClim 2, Harmonized World Soil Database and Shuttle Radar Topography Mission (SRTM), respectively. The models presented good performance with coefficient of determination (R2) values ranging from 0.68 to 0.84, mean absolute error (MAE) between 0.32 and 0.96%.year−1, and root mean square error (RMSE) ranging from 60 a 160%. The recruitment was higher than mortality, and the basal area gain was greater than the basal area loss, resulting in positive net change rates. The most important variables for the model and best predictors of vegetation dynamics rates of Brazilian savanna were the terrain variables, since they encompass characteristics such as soil water content, erosion, altitudes and slope. We showed that the gain in number of individuals and basal area were greater at low regions, that is, near drainage lines and with higher humidity. The losses in number of individuals and basal area were higher in boundary regions of the savanna vegetation, that is, regions of higher altitude and consequently, low humidity. Therefore, our findings provide support for conservation and management strategies of the biome, since from the vegetation dynamics drivers identification, we may identify priority areas as well as the vegetation vulnerability.|
|Appears in Collections:||DCF - Artigos publicados em periódicos|
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.