the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Mar 2026
Linking individual-based forest modelling with a radar simulator for determining forest structure and biomass
Abstract. Mapping forest structure, critical for assessing carbon stocks and fluxes, remains challenging with remote sensing. We propose a novel framework linking an individual-based forest model (FORMIND), which generates explicit 3D forest structures and dynamics, with a radar simulator (here used for TanDEM-X). We investigate radar coherence from simulated forests to predict aboveground biomass (AGB) across varying spatial scales, measurement noise levels, and successional stages. The framework is applied to the Barro Colorado Island (BCI) tropical forest, where we evaluate simulated coherence against TanDEM-X observations and invert canopy height, comparing the results with airborne laser scanning (ALS) data.
Results indicate a positive link between forest structure and interferometric patterns, with AGB prediction showing a clear dependence on spatial resolution. This novel approach offers a pathway to map forest structure by combining broad radar data coverage with an ecologically explicit forest model.
- Preprint
(2664 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-613', Anonymous Referee #1, 23 Mar 2026
-
RC2: 'Comment on egusphere-2026-613', Anonymous Referee #2, 21 Apr 2026
Dear authors,
the overall research topic addressed in this paper is certainly relevant, and especially the insights of a 4D forest structure model to better parameterize and analyse a decorrelation model from radar observations.
Nonetheless, the paper suffers from several drawbacks and its content is rather disappointing regarding the proposed research questions.
First of all, the so-called 'radar simulator' lies actually just on an elegant but simple analytical formulation of the volume decorrelation proposed 20 years ago, which adaptation using the Leaf Area Density is clearly not sufficiently justified. Indeed, we would expect much more from a 'radar simulator' (with for instance speckle modeling, non homogeneous 3D representation, etc), and detailed explanations of how the extinction term inside the integral is replaced by the LAD (although correct mathematically, the physical links would be significantly more complicated). In that respect, please note that none of the cited references ([Treuhaft,1996 or 2000]) give this shortcut (as the writing could let us understand).
Likewise, the paper also reduces forest structure to the LAD (vertical profile of LAI), which could be also disappointing for the reader.
Regarding then the methodology, I really do not see why forest AGB and height are not analyzed together, using the same criteria (it seems for instance that the noise model would not impact forest height retrieval, or that biomass retrieval would not be relevant or worth to show). This is particularly regrettable since the FORMIND model would give a very detailed (in terms of vegetation component) link between both.
Regarding the results, the sensitivity of forest AGB to coherence amplitude is clearly overstated, but it reveals the limitations of the analytical model for volume decorrelation. Height retrieval seems much better that what could be expected at X-band, but the details given in §2.6 are not sufficient to really understand what has been done, especially with the use of inventory data (l138, no more). I would therefore strongly recommend to refocus the paper on this latter part (on forest height inversion), with further details on the processing steps and comparison with standard methods that do not exploit the FORMIND LAD information (to better emphasize its importance through this formulation).Citation: https://doi.org/10.5194/egusphere-2026-613-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 128 | 48 | 16 | 192 | 10 | 17 |
- HTML: 128
- PDF: 48
- XML: 16
- Total: 192
- BibTeX: 10
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript with reference ID egusphere-2026-613 presents a novel framework linking an individual-based forest model (FORMIND) with a radar simulator (TanDEM-X) with the goal to investigate the relationship between forest structure, interferometric coherence, and aboveground biomass (AGB). The approach is applied to the Barro Colorado Island (BCI) tropical forest, where simulated radar signals are evaluated against observations and canopy height estimates are compared with airborne laser scanning (ALS) data. The study demonstrates that forest structural properties influence radar coherence and that AGB retrieval is scale-dependent, highlighting the potential of combining ecological modelling with radar remote sensing.
General comments:
This study presents an innovative and interdisciplinary study bridging forest ecology, remote sensing, and modelling. However, the manuscript would benefit from clearer and more concise presentation of the methodological limitations and the transferability of the study finding. For instance, the authors indicate that the applied method "is transferable via ecological parameterization and directly relevant for missions such as ESA BIOMASS". However the manuscript (as it currently stands) is lacking actual examples of how to make us of the proposed approach for for early warning change detection or improved global carbon accounting. Most strikingly, the manuscript closes with a remark on improving robustness across successional stages of forests and the appendix (Figure A1) actually highlights some of the accuracy across successional stages and noise levels for tropical forest in Panama, BCI. Hence, to me this appears to be one of the most interesting results that should find their way into the main text, which currently just presents a rather technical description of the applied framework and thus could be improved by a more lively discussion of the study findings in light of the recent literature.
Specific comments:
Figure 4 depicts the results of biomass prediction from different radar coherence metrics but it remains unclear why the slopes change from negative (in panels a, b) to positive (in panels c, d).
Figure 5 shows forest height and radar coherence but what are the units of y-axis (in panel a, b)?
Figure A1 indicates the accuracy of biomass prediction across different successional stages, which to me is one of the most interesting results as this highlights a shift of biomass (and its accuracy) across early-, mid-, and late-, succession, which could be added to the main text.