Assessing the effects of uncertainty in remote sensing-based predictions of tree species composition and site index on net present value calculations

Abstract Accurate prediction of forest attributes such as species composition and site index (SI) is critical for effective forest management and decision-making. Uncertainties in these attributes can lead to suboptimal management decisions and thus economic losses. In this study, rule-based simulations were used to calculate net present value (NPV) as an economic indicator to assess the relative impacts of uncertainty in species composition and SI predictions. Species composition and SI were modeled and predicted using airborne laser scanning data and multispectral imagery data from Senitnel-2. In addition, the benefits of using calibration to improve the accuracy of species composition and SI predictions were evaluated. The objectives of the study were: (i) to investigate whether errors in species composition or SI had a greater impact on NPV calculations, (ii) to assess the effect of calibration on prediction accuracy, and (iii) to evaluate the combined effects of uncertainties in species composition and SI on NPV calculations. Monte Carlo cross validation and paired bootstrap were used to assess prediction performance and NPV impacts on training and independent validation data, respectively. Uncertainty of calculated NPV ranged between 880 and 1571€/ha (6.7%–12%) in cross validation and from 284 to 1304€/ha (3.6%–16.4%) in independent validation. Uncertainty in SI had a greater negative impact on NPV calculations than uncertainty in species composition. Calibration of species composition and SI predictions improved NPV calculations, particularly in independent validation. The combined use of predicted species composition and SI resulted in the largest errors in NPV calculations. These findings emphasize the need to reduce species composition and SI uncertainty and refine automated methodologies to improve the accuracy of forest economic assessments and decision-making.