Tree biomass plays a key role in sustainable management by providing different aspects of ecosystem. Estimation of above ground biomass by non-destructive means requires the development of allometric equations. Most researchers used DBH (diameter at breast height) and TH (total height) to develop allometric equation for a tree. Very few species-specific allometric equations are currently available for shrubs to estimate of biomass from measured plant attributes. Therefore, we used some of readily measurable variables to develop allometric equations such as girth at collar-height (GCH) and height of girth measuring point (GMH) with total height (TH) for A. rotundifolia, a mangrove species of Sundarbans of Bangladesh, as it is too dwarf to take DBH and too irregular in base to take Girth at a fixed height. Linear, non-linear and logarithmic regression techniques were tried to determine the best regression model to estimate the above-ground biomass of stem, branch and leaf. A total of 186 regression equations were generated from the combination of independent variables. Best fit regression equations were determined by examining co-efficient of determination (R2), co-efficient of variation (CV), mean-square of the error (MSerror), residual mean error (Rsme), and F-value. Multiple linear regression models showed more efficient over other types of regression equation. The performance of regression equations was increased by inclusion of GMH as an independent variable along with total height and GCH. Estimation of above ground biomass by non-destructive means require the development of allometric equations. Most researchers used DBH (diameter at breast height) and TH (total height) to Very few species-specific allometric equations to a tree. Very few species-specific allometric equations to a tree. Very few species-specific allometric equations to a all-scale equations such as girth at collar-height (GCH ) and height of girth measuring point (GMH) with total height (TH) for A. rotundifolia, a mangrove species of Sundarbans of Bangladesh, as it is too dwarf to take DBH and too irregular in base to take Girth at a fixed height. Linear, non-linear and logarithmic regression techniques were tried to determine the best regression model to estimate the above-ground biomass of stem, branch and leaf. A total of 186 regression equations were generated from the combination of independent variables. Best fit regression equations were determined by examining co-efficient of determination (R2), co-efficient of variation (CV), mean-square of the error (MSerror), residual mean error (Rsme), and F-value. Multiple linear regression models showed more efficient over other types of regression equations. The performance of regression equations was increased by inclusion of GMH as an independent variable along with total height and GCH.