Quality information on forest resources is fundamental for sustainable forest management. Manual aerial photointerpretation is used as a cost-effective source of data for forest inventories; however, the process of photointerpretation is inherently subjective and is often undertaken by multiple photointerpreters for a given forest management area. In contrast, airborne laser scanning (ALS) data enable characterization of forest structure in a systematic fashion with quantifiable levels of accuracy and precision that often exceed required targets and standards. However, the gains associated with the use of new technologies for forest inventory are difficult to measure because the quality of existing photointepreted inventories have rarely been quantified. Using ALS data as reference, the objective of this study was to quantify the precision of photointerpreted estimates of forest stand height and canopy cover (CC). We examined forest inventories from three study sites in three different forest regions of Canada. Each of the study sites was located within a different provincial jurisdiction with unique photointerpretation standards and forest ecosystems. Stand-level estimates of forest height and cover were compared to reference estimates generated from the ALS data. Overall, our results indicated that precision was greater for photointerpreted estimates of height, with a relative standard deviation ranging from 22 per cent to 29 per cent among our three sites, compared to estimates for CC, with precision ranging from 28 per cent to 59 per cent. While the relationship between photointerpreted estimates of height and ALS estimates of height were generally linear and consistent for all study sites, relationships for CC were non-linear. We found that precision for both stand height and cover varied by dominant species, inventory stand structure, age, and ALS canopy complexity, and that in the majority of cases, the differences between the photointerpreted estimate and the ALS estimate were statistically significant. It is also noted that the variability in photointerpretation precision as a function of the aforementioned factors was not consistent among our three study sites, indicating that site-specific forest conditions and photointerpretation procedures influence the precision of photointerpreted estimates. The influence of local forest conditions and interpretation procedures are therefore important considerations when seeking to quantify the potential relative gains in precision, which may be afforded by technologies such as ALS for forest inventory programs. Moreover, approaches to improve consistency in photointerpreted estimates of cover would be useful for operational inventory programs.

中文翻译：

---

量化从航空照片解释得出的林分高度和冠层覆盖估计的精度

森林资源的质量信息是可持续森林管理的基础。人工航空照片判读被用作森林清查的经济有效的数据来源；然而，照片翻译过程本质上是主观的，并且通常由多个照片翻译员针对给定的森林管理区域进行。相比之下，机载激光扫描 (ALS) 数据能够以系统的方式对森林结构进行表征，其准确度和精确度的可量化水平通常超过所需的目标和标准。然而，与使用新技术进行森林清查相关的收益很难衡量，因为现有的光解清查存的质量很少被量化。使用 ALS 数据作为参考，本研究的目的是量化光解估计的森林高度和树冠覆盖 (CC) 的精度。我们检查了加拿大三个不同森林地区三个研究地点的森林清单。每个研究地点都位于不同的省辖区内，具有独特的照片解释标准和森林生态系统。森林高度和覆盖的林分水平估计值与 ALS 数据生成的参考估计值进行了比较。总体而言，我们的结果表明，照片解释的高度估计精度更高，我们三个地点的相对标准偏差范围为 22% 至 29%，而 CC 的估计精度为 28% 至 59%。分。虽然照片解释的身高估计和 ALS 身高估计之间的关系在所有研究地点通常是线性的并且是一致的，但 CC 的关系是非线性的。我们发现，林分高度和覆盖度的精度因优势物种、库存林分结构、年龄和 ALS 冠层复杂性而异，并且在大多数情况下，光解估计和 ALS 估计之间的差异具有统计学意义。还值得注意的是，作为上述因素的函数的光解精度的变异性在我们的三个研究地点之间并不一致，这表明特定地点的森林条件和光解程序会影响光解估计的精度。因此，在寻求量化潜在的相对精度增益时，当地森林条件和解释程序的影响是重要的考虑因素，这可能由森林清查计划的 ALS 等技术提供。此外，提高照片解释覆盖率估计的一致性的方法对于运营清单计划很有用。