Constructing a non-linear additive crown-width model system for moso bamboo forests in eastern China

Abstract

ABSTRACT Tree crown can be described by various crown dimension variables, including crown width (CW), an important predictor variable in forest growth-and-yield models. Because CW is quantified as the arithmetic mean of north–south and east–west CW measurements (crown components), issues of additivity and inherent correlation between crown components exist. Using two versatile estimation algorithms (non-linear seemingly unrelated regression − NSUR – and non-linear error-in-variable models − NEIVM), we employed three methods of constructing simultaneous model systems (non-linear summation − NSE, one-step proportional weighting system − OPWS, and two-step proportional weighting system − TPWS) to effectively address those issues. The performance of the three methods was evaluated using CW measurements of moso bamboo (Phyllostachys edulis) in China. Results showed that the OPWS outperformed the NSE and the TPWS for each algorithm. NEIVM-OPWS was more accurate than TPWS, and NEIVM outperformed NSUR. Diameter at breast height, total height and mean diameter at breast height (a variable describing stand density and competition) significantly contributed to CW and its component equations. Each CW model system included additivity of the crown component equations and accounted for the inherent correlations between equations. Overall, NEIVM-OPWS provided a closer estimate of CW to the measured CW than other model systems and methods employed. An effective CW model system will help understand light capture by the bamboo canopy and contribute to the effective management of bamboo forests in China and elsewhere.