LOG-VEROSSIMILHANÇA COMBINADA PARA COMPARAÇÃO DE ESTRUTURAS DE CONTINUIDADE ESPACIAL EM FLORESTA ESTACIONAL DECIDUAL

The present study aimed to examine spatial dependence and distribution of volume in three fragments of Subtropical forest, geographically separated and the different post-intervention ages, using a single geostatistical model or combined model. Data were collected from 56 sampling units of 250 m 2 systematically distributed in a grid of 40 x 40 m. Trees with DBH ≥ 10 cm were measured according to Prodan’s methodology. Two types of analysis were applied to the data. The first one was an individual adjustment for comparison between their models and the second one consisted in the proposed combined approach adjustment. Both analysis used geostatistical models with adjustment function maximizing log-likelihood. Models were compared using Akaike criterion (AIC) and relational range parameter as an indication of spatial dependence degree. Results show that combined models had lower AIC values as well as greater spatial dependence degree on adjustments of individual areas models. This research indicates the use of combined log-likelihood geostatistical models to study fragmented forests for analysis and detection of spatial volume correlation structure.