Adapting distance-independent forest growth models to represent spatial variability : Effects of sampling design on model coefficients