the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Mar 2026
Why multitemporal ALS forest metrics remain challenging: Insights from operational airborne laser scanning
Abstract. The increasing availability of large-scale and repeated operational airborne laser scanning (ALS) facilitates its use for multitemporal analyses and monitoring. In the last decades, single ALS acquisitions have been demonstrated to hold great potential for obtaining forest structural information. However, the robustness and reliability of ALS to accurately detect changes in complex forest structural parameters such as plant area index (PAI), from repeated ALS acquisitions have rarely been assessed. In this study, we evaluated the reliability and limitations of multitemporal mapping and interpretation of this structural trait, using the well-established canopy height (CH) as a reference metric. We used operational ALS data from a heterogeneous temperate forest in northern Switzerland from three years, 2014, 2019 and 2020, recorded with different sensor and flight settings. Our results showed that CH was largely unaffected by the differences in data acquisition, reaffirming it to be a robust trait and demonstrating that our data is usable for multitemporal analyses. For PAI, we applied and compared three estimation methods with varying complexity. PAI results were highly sensitive to various acquisition parameters, particularly the pulse repetition frequency, leading to large deviations between acquisitions. All tested PAI estimation methods exhibited similar problems, complicating the distinction between actual structural change and external effects. This study underscores the potential of operational ALS for multitemporal forest structure analyses but highlights the need for standardisation of recording parameters as much as possible, as well as methodological harmonization and calibration to ensure comparability in multitemporal analyses, particularly for complex forest structural traits.
- Preprint
(1875 KB) - Metadata XML
-
Supplement
(581 KB) - BibTeX
- EndNote
Status: open (until 13 May 2026)
- RC1: 'Comment on egusphere-2026-1689', Anonymous Referee #1, 04 May 2026 reply
-
RC2: 'Comment on egusphere-2026-1689', Anonymous Referee #2, 04 May 2026
reply
The authors assess whether repeated operational airborne laser scanning (ALS) acquisitions, collected with different sensors and flight settings, can be used reliably for multitemporal forest structure analysis. They use ALS data acquired over a heterogeneous temperate forest in Switzerland in 2014, 2019, and 2020, and compare three PAI estimation methods of varying complexity: leafR, the Arnqvist method, and AMAPVox. This is an important topic, as the ability to derive robust structural metrics across different data sources, sensors, and acquisition settings is increasingly relevant given the growing availability of ALS datasets.
However, I have two major concerns about the current manuscript. First, the interpretation of interannual PAI differences relies strongly on the assumption that the selected reference forests were structurally stable over time. The reference forest is defined using conifer dominance, tall canopy height, and small H95 change between 2014 and 2020. While this may indicate limited change in upper-canopy height, it does not necessarily demonstrate stability in canopy openness, or vertical structural heterogeneity. This is important because the manuscript itself distinguishes CH and PAI as representing different dimensions of canopy structure. Additional structural indicators and/or independent disturbance and management information are therefore needed to support the “no-change” assumption.
Second, I am not yet fully convinced that the manuscript provides a sufficiently clear advance beyond confirming known limitations of ALS-derived density metrics. It is well recognised that PAI/PAD and related vertical-density structural metrics are sensitive to canopy occlusion, pulse penetration, pulse density, scan angles and other sensor settings, particularly in lower canopy layers. Demonstrating that repeat operational ALS acquisitions produce inconsistent PAI estimates is useful, but the manuscript should more clearly explain what new knowledge is gained relative to previous work. If the authors do not aim to propose a correction or harmonisation method, the study would benefit from providing stronger quantitative diagnostics or transferable guidance, for example by identifying acquisition or return-distribution conditions under which ALS-derived PAI can or cannot be interpreted reliably.
Overall, the authors provide a useful operational case study, I appreciate their substantial processing effort and the practical relevance. However, the study would be strengthened by more rigorously validating the structural stability of the reference forests, more cautiously separating acquisition effects from real forest dynamics, and more clearly defining the novelty and transferable contribution of the work.
Specific comments are provided below.
1. Reference forest definition and the “no structural change” assumption:
The interpretation that PAI differences are largely driven by acquisition settings is plausible, but the current design does not fully separate acquisition effects from real forest dynamics, management, or drought-related structural change. Without stronger evidence that the reference areas were structurally stable, it remains difficult to determine whether interannual PAI differences reflect acquisition effects, methodological effects, or actual changes in forest structure.
The authors assume PAI should remain relatively consistent across years in structurally stable forest patches, and that interannual deviations therefore mainly reflect acquisition-related or methodological effects. The reference forest is selected using conifer dominance, tall canopy height, and small H95 change between 2014 and 2020. These criteria identify tall coniferous pixels with limited change in upper-canopy height, but stability in H95 does not necessarily imply stability in canopy openness, or vertical structural heterogeneity. H95 mainly characterises the upper canopy and miss changes in canopy cover, gap fraction, and within-canopy complexity. I therefore recommend that the authors evaluate the reference areas using additional structural attributes beyond H95. Useful and relatively simple metrics could include mean canopy height, canopy cover or gap fraction, height variability such as height standard deviation and coefficient of variation, and a standardised rumple index. These metrics have been shown in previous studies to be relatively robust across ALS data sources and acquisition settings, and together they would better characterise the multivariate nature of forest structure. Using a set of simple, robustly retrievable metrics would help demonstrate whether the selected reference areas were truly stable in structural dimensions relevant to PAI, rather than only in top-canopy height.
Independent disturbance and management information would also be valuable. Records of thinning, harvesting, windthrow, drought damage, insect outbreaks, disease, or other disturbances in the study area during 2014–2020 could be used to validate the assumption that the reference forest experienced little structural change. This is particularly important because the Discussion mentions drought years and possible local canopy decline or mortality, but the reference forest and PAI comparisons do not appear to have been screened using independent disturbance or management records.
2. Attribution of PAI differences to acquisition parameters needs stronger quantitative support
The interpretation of acquisition effects, particularly the role of pulse repetition frequency, should be made more cautious. Several acquisition parameters changed simultaneously between years, including sensor type, flight altitude, scan angle, point density, beam divergence, PRF, and potentially sensor-specific receiver settings or processing differences. Therefore, the current study design cannot fully isolate the effect of PRF from other acquisition-related factors.
The explanation that higher PRF in 2019 reduced pulse energy and led to fewer intermediate returns is plausible. The example patches shown in Figure 7 are useful for illustrating this mechanism, but it is unclear how representative these patches are. However, the evidence currently appears to be based mainly on selected patch-level comparisons rather than a statistical analysis. As a result, the conclusion in the ‘Influence of sensor parameters’ section is somewhat over-interpreted.
3. Clarify whether PAI method comparisons are based on the reference forest or the broader landscape
The manuscript states that the reference forest was intended for PAI method comparison. However, several of the main method-comparison results appear to be presented for the broader study area rather than only for the reference forest, for example Figures 4–6. The authors should clearly specify which spatial subset was used to generate each result: the full forested study area, the reduced valid PAI tile area, the reference forest, or forest-type-specific subsets.
If the PAI method comparison is based on the whole landscape or broader PAI analysis area, then forest structure cannot necessarily be assumed to have remained static across the three acquisition years. In that case, real forest dynamics, management activities, drought effects, and forest-type differences could also contribute to interannual PAI differences. The authors should therefore explain the rationale for using the broader landscape for method comparison and discuss how potential structural change was accounted for. Conversely, if the goal is to assess which PAI method is most temporally consistent under stable forest conditions, then the reference forest should be central to the comparison.
4. Broader ecological implications and applications need stronger discussion
The Discussion currently focuses mainly on methodological and sensor-related explanations for the observed variation in ALS-derived PAI. This is useful and relevant, however, the broader ecological and operational implications of the findings are not sufficiently developed. The Introduction motivates the study in terms of forest structure, ecosystem functioning, biodiversity, and forest monitoring, but the Discussion does not fully return to these themes.
Minor comments
L 45: Please clarify the definition of LAI, as was done for PAI, and more explicitly explain the distinction between LAI and PAI. Since this study estimates PAI from leaf-off ALS data rather than LAI, the authors should make clear that the derived metric likely includes woody material and is not directly comparable to true leaf area, especially in deciduous forests.
L155: Please clarify why a 2 m resolution was selected, both methodologically and ecologically.
L165: Hmax and H95 should be calculated using first returns. Authors need to have a check and clarify this.
Figure 2a appears to include 0.5 m, whereas the Methods mention 1, 2, 5, 10, and 20 m.
Figures 3–6: Please state clearly whether the results shown in Figures 3–6 were derived at 2 m resolution, and clarify the spatial domain used for each figure, for example the full study area, reduced PAI analysis area, reference forest, or forest-type subsets.
Figure 4: Please consider using the same x-axis scale and breaks for the PAI density plots to make comparison across the three methods more intuitive.
Citation: https://doi.org/10.5194/egusphere-2026-1689-RC2
Data sets
ALS dataset 2020 swisstopo https://www.swisstopo.admin.ch/de/hoehenmodell-swisssurface3d
ALS dataset 2014 Canton Aargau https://www.ag.ch/de/themen/staat-politik/daten-und-zahlen/geoportal/geodaten/geodatenliste?rewriteRemoteUrl=/details/AGIS.kai_lidarpoint14u?searchcontext%3DLidar
ALS dataset 2019 Canton Aargau https://www.ag.ch/de/themen/staat-politik/daten-und-zahlen/geoportal/geodaten/geodatenliste?rewriteRemoteUrl=/details/AGIS.kai_lidarpoint19?searchcontext%3DLidar
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 183 | 74 | 18 | 275 | 37 | 11 | 16 |
- HTML: 183
- PDF: 74
- XML: 18
- Total: 275
- Supplement: 37
- BibTeX: 11
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
overall the manuscript is interesting and well structured. The results are interesting for the scientific community.
The presentation quality - results are presented good.
Major issue: In the discussion I miss in general comparison of your results with literature. Also, in explaining your results are references to other studies missing. From my point of view the discussion needs improvement.