Tree height-diameter allometry and implications for biomass estimates in Northeastern Amazonian forests

The relationship between tree height and diameter varies across forest types, introducing uncertainties in height that can affect aboveground biomass estimates in tropical forests. Here, we used a four-step approach to assess whether incorporating height estimates from local height-diameter models, compared to two published equations, improves biomass estimates across spatial scales. First, we measured the diameter and height of 1,962 trees in two representative forest types in the Northeastern Amazon: non-flooded terra-firme and seasonally-flooded várzea forests. Second, we selected the best height-diameter models from a set of 10 candidates to establish local allometric equations. Third, we applied these best local models and two previously published height models (the regional Guyana shield, and the pantropical model) to estimate tree height, and compared these estimates to measured height. Finally, we computed tree biomass using equations that both included and excluded height, and compared these biomass estimates to those calculated using directly measured height. Asymptotic height-diameter models provided the best fit at local and regional scales. The Quadratic model was the best choice for terra-firme and várzea forests separately, while the Weibull and Michaelis-Menten models performed best for both forests. Local models closely matched measured heights, with deviations of only 0.1%, outperforming the regional and pantropical models within each forest type. The regional model underestimated height in terra-firme by 3% and overestimated it in várzea by 29%, while the pantropical model underestimated height in terra-firme by 19% and overestimated it in várzea by 6%. Using local asymptotic models to estimate height improved the accuracy of biomass estimates, with differences of around 1% between biomass computed using measured heights and estimated heights for terra-firme and várzea forests. In contrast, the biomass calculated using estimated heights from both the regional and pantropical models overestimated the biomass in várzea by 41% and 17%, respectively, while the pantropical model underestimated biomass in terra-firme by 17%. The estimated height and biomass of large trees using regional and pantropical models showed the highest deviations from the observed values. Our findings underscore the necessity for height-diameter modeling for different forest types, and highlight the need to increase sampling of large trees to improve biomass estimation accuracy in Northeastern Amazonia.