We examined the accuracy of the stand attribute data based on airborne laser scanning (ALS) provided by the Finnish Forest Centre. The precision of forest inventory data was compared for the first time with operative logging data measured by the harvester. Airborne laser scanning (ALS) is increasingly used together with models to predict the stand attributes of boreal forests. The information is updated by growth models. Information produced by remote sensing, model prediction, and growth simulation needs field verification. The data collected by harvesters on logging sites provide a means to evaluate and verify the accuracy of the ALS-based data. This study investigated the accuracy of ALS-based forest inventory data provided by the Finnish Forest Centre at the stand level, using harvester data as the reference. Special interest was on timber assortment volumes where the quality reductions of sawlog are model predictions in ALS-based data and true realized reductions in the logging data. We examined the accuracy of total volume and timber assortment volumes by comparing ALS-based data and operative logging data measured by a harvester. This was done both for clear cuttings and thinning sites. Accuracy of the identification of the dominant tree species of the stand was examined using the Kappa coefficient. In clear-felling sites, the total harvest removals based on ALS and model prediction had a RMSE% of 26.0%. In thinning, the corresponding difference in the total harvested removal was 42.4%. Compared to logged volume, ALS-based prediction overestimated sawlog removals in clear cuttings and underestimated pulpwood removals. The study provided valuable information on the accuracy of ALS-based stand attribute data. Our results showed that ALS-based data need better methods to predict the technical quality of harvested trees, to avoid systematic overestimates of sawlog volume. We also found that the ALS-based estimates do not accurately predict the volume of trees removed in actual thinnings.

中文翻译：

---

根据实际测井数据评估基于 ALS 的清除估计的准确性

我们基于芬兰森林中心提供的机载激光扫描 (ALS) 检查了林分属性数据的准确性。首次将森林清查数据的精度与采伐机测量的操作伐木数据进行了比较。机载激光扫描 (ALS) 越来越多地与模型一起用于预测北方森林的林分属性。该信息由增长模型更新。遥感、模型预测和生长模拟产生的信息需要现场验证。采伐者在伐木场收集的数据提供了一种评估和验证基于 ALS 数据准确性的方法。本研究使用收割机数据作为参考，调查了芬兰森林中心在林分级别提供的基于 ALS 的森林清单数据的准确性。特别感兴趣的是木材分类量，其中锯木的质量降低是基于 ALS 数据的模型预测和测井数据中真正实现的减少。我们通过比较基于 ALS 的数据和采伐机测量的操作测井数据，检查了总体积和木材分类体积的准确性。这样做是为了清除切割和变薄现场。使用 Kappa 系数检查林分优势树种识别的准确性。在未采伐地点，基于 ALS 和模型预测的总采伐量的 RMSE% 为 26.0%。在间伐中，总收获去除量的相应差异为 42.4%。与测井体积相比，基于 ALS 的预测高估了清晰插条中的锯木清除量，而低估了纸浆木材的清除量。该研究为基于 ALS 的林分属性数据的准确性提供了有价值的信息。我们的结果表明，基于 ALS 的数据需要更好的方法来预测采伐树木的技术质量，以避免系统地高估锯木量。我们还发现，基于 ALS 的估计不能准确预测实际间伐中移除的树木数量。