Search

Earthworm biomass is one of the most important quantitative indices of forests soil. Estimating the earthworm biomass in forest soils is often difficult due to changes in soil properties. This study aims at comparing the performance of multiple linear regression and regression tree models in estimating the earthworm biomass in different forest conditions. Earthworm biomass was sampled in 40 forest gaps and adjacent virgin forest. They were then separated from soil using hand-sorting method and weighted in 0.01 gr precision in compartment 32 Shastkolate forest. Physical and chemical soil properties were analyzed by using standard laboratory methods. The best fitted models were specified in multiple linear regression and regression tree models for estimating the earthworm biomass. The fitted models were then validated by using Mean Error (ME), Root Mean Square Error (RMSE) and Relative Error (RE) measures. Soil bulk density and total nitrogen explained 24% of total earthworm biomass variances in virgin forest while canopy area solely explained 66% of total variances in canopy gap. Multiple linear regression models overestimated the earthworm biomass while regression tree model underestimated the biomass in virgin forest but it was precise estimator in canopy gap. According to the results and due to heterogeneity in forest environment, it is recommended the research is done hierarchically for finding the important and effective variables (gap area and soil potassium in the current study) and after the site stratification based on these variables, the relationship between independent and dependent variables is studied.