This study investigated forest cover change and the driving forces behind it in Fagita Lekoma District of Ethiopia that resulted in increased forest cover, which might be uncommon outside this case study area. The LULC change analysis was made from 2003 to 2017 based on Landsat images. Socioeconomic analysis was carried out to identify the major driving forces that resulted in LULC change. A questionnaire survey, focused group discussion, key informant interviews and field observation were employed to analyze the link between LULC change and the driving forces. The 15-year period (2003–2017) image analysis revealed that the coverage of forest lands, built-up areas and grassland has increased by 256%, 100% and 96%, respectively, at the expense of cultivated lands and wetlands. The increased forest cover is due to the woodlots expansion of Acacia decurrens Willd, which are designed for sustainable livelihoods and a land revitalization strategy in the study area. Rapid population growth, an increasing demand for charcoal and subsequent market opportunities, preferred qualities of A. decurrens or black wattle to halt land degradation as well as to improve land productivity, have been identified as the major driving forces of forest cover change. Chi squared analysis revealed that: a comparative cash income from the sale of A. decurrens; a dependency on natural forests; the distance from the district administrative center; the size of the active labor force, and the area of land owned have significantly affected the cover change. The major forest cover change is due to the expansion of A. decurrens plantations that have socioeconomic and environmental implications to improve rural livelihoods and revitalize the land. Thus, the positive experiences identified in this study should be scaled-up and applied in other similar settings.

Bivariate distribution models are veritable tools for improving forest stand volume estimations. Their accuracy depends on the method of construction. To-date, most bivariate distributions in forestry have been constructed either with normal or Plackett copulas. In this study, the accuracy of the Frank copula for constructing bivariate distributions was assessed. The effectiveness of Frank and Plackett copulas were evaluated on seven distribution models using data from temperate and tropical forests. The bivariate distributions include: Burr III, Burr XII, Logit-Logistic, Log-Logistic, generalized Weibull, Weibull and Kumaraswamy. Maximum likelihood was used to fit the models to the joint distribution of diameter and height data of Pinus pinaster (184 plots), Pinus radiata (96 plots), Eucalyptus camaldulensis (85 plots) and Gmelina arborea (60 plots). Models were evaluated based on negative log-likelihood (−ΛΛ). The result show that Frank-based models were more suitable in describing the joint distribution of diameter and height than most of their Plackett-based counterparts. The bivariate Burr III distributions had the overall best performance. The Frank copula is therefore recommended for the construction of more useful bivariate distributions in forestry.

The objective of this work was to compare estimates generated by a diametric distribution model and a total stand model against the pre-cut inventory. The model efficiency was also evaluated. Data were evaluated from 30 permanent sample plots in a Eucalyptus urophylla stand, comprising 24 sample plots used for model fitting, and six sample plots for validation. The volume of wood per hectare was estimated for different productive units (sites), using 7 years as the reference age. The model adjustment quality was verified by adjustment and precision statistics: the correlation between observed and predicted variables, root mean square error percentage, graphical analysis of residual distribution, and a frequency histogram for classes of relative errors and validation. Although the two-parameter Weibull probability density function adhered to the data for tree evolution in diameter classes for the reference age (7 years) in the different productivity classes, it generated imprecise estimates of the number of individuals. Consequently, it produced inaccurate volumetric production estimates. The total stand model provided reliable projections of production volumes in different productivity classes for both adjustment types, showing compatibility with the pre-cut inventory according to a Tukey test. In summary, the total stand model generated estimates that were compatible with the pre-cut inventory while the diametric distribution model did not.