the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 May 2025
Litter vs. Lens: Evaluating LAI from Litter Traps and Hemispherical Photos Across View Zenith Angles and Leaf Fall Phases
Abstract. Leaf area index (LAI) is a key parameter for modeling ecosystem productivity, climate interactions, and hydrological processes, as well as monitoring vegetation health. While satellite-based estimates provide insights into large-scale vegetation dynamics, ground-based methods, including digital hemispherical photography (DHP), are essential to generate and validate such products and offer a practical alternative for fine-scale assessments. However, it remains unclear if the DHP method enables to robustly track temporal LAI dynamics. Here, we evaluate DHP-derived LAI time series with litter trap (LT)-derived LAI in a temperate deciduous broad-leaved forest. First, by comparing DHP-derived LAI estimates with LT-derived LAI across varying view zenith angles ranging from 10° to 90°, we investigate how well both methods align. Using 15 sample locations, we found the highest average correlation across all locations of DHP- and LT-derived LAI (R2 =0.88) at a view zenith angle of 20°, indicating that litter traps represent a relatively narrow spatial footprint. Uncertainties for individual litter traps attributed to varying site conditions, such as tree stem density or canopy coverage. To overcome these uncertainties, we applied a site specific calibration using the litter traps and a generalized linear mixed model, which significantly increased correlation (R2 =0.97).
This study highlights the potential of DHP for tracking spatio-temporal LAI dynamics in decideous forests. Moreover, we demonstrate that integrating DHP and LT data, alongside a mixed-effects model, can enhance the site specific accuracy and applicability of LAI assessments.
- Preprint
(15865 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
CC1: 'Comment on egusphere-2025-1496', Hongliang Fang, 26 May 2025
reply
-
AC1: 'Reply on CC1', Simon Lotz, 03 Jun 2025
reply
Dear Prof. Hongliang Fang,
Thank you for your valuable comments, particularly for highlighting additional literature and relevant citations.
We greatly appreciate your insights and will definitely address these points in the revised version of the paper.Citation: https://doi.org/10.5194/egusphere-2025-1496-AC1
-
AC1: 'Reply on CC1', Simon Lotz, 03 Jun 2025
reply
-
RC1: 'Comment on egusphere-2025-1496', Francesco Chianucci, 24 Jun 2025
reply
The manuscript compared temporal variation of LAI, as measured from the semi-direct benchmark littertrap method, with DHP, using different clumping corrections. The relative novelty of the study is to use temporal series, which allows to investigate how the accuracy between methods varied with canopy development (and thus LAI density).
The study focused on beech forests, which are very dense, and where the recent LXG method can correct more for clumping index, compared with other approaches. The result agrees with previous finding indicating that LXG is more suited for nearly non-transparent crowns.
The main issues in my view are about
1) data collection. Fisheye photos are strongly influenced by sky condition and camera exposure. The authors collected data in direct sky conditions and likely automatic exposure. Both options greatly influence gap fraction retrieval, with generally increase the underestimation of LAI from DHP. This is apparent in the sample images in Fig2-Fig3 where the sky is likely saturating (clipping) and this can largely overestimate gap fraction (and thus underestimate LAI). Additionally, gamma correction is another factor required to provide linear relationship between image brightness and radiance.
There is a wide range of literature confirming this, and therefore the accuracy of the method can be biased compared to litter traps. This should be acknowledged and authors could try to find some solution to this issue. For example, if they collected RAW imagery, they can try to analyse raw images using the bRaw R package, https://www.biorxiv.org/content/10.1101/2022.10.25.513518v1, which resemble a method by Macfarlane et al. 2014, to linearly scale 12 to 8 bit raw blue imagery to reduce DHP sensitivity to camera exposure.
2) The DHP invert LAI without requiring information on extinction coefficient by integrating multiple gap fraction (GF) measurements over the full zenith range, as the integral of G(theta) is 0.5 over 0-90°, as per the Miller 1967 theorem. By reducing the zenith angle view, the LAI inversion is not straightforward. Therefore, the higher correlation with LT at smaller VZA is in my view the results of higher clumping correction, as the inner rings could have more variability, which reduced the underestimation of DHP. Authors should de-structure their LAI component obtained from DHP, namely explore the variation in GF, CI and G(theta) (here using LAI from LT) across VZA, to explore how these three attributes varied with VZA. This would shed some light on the different performance of canopy units like GF, G and CI to the resulting LAI estimation from DHP.
Specific comments:
Direct, semi-direct vs indirect methods
LT is typically considered a semi-direct method (Bréda 2003), in that the leaf area is directly measured, as for the allometry methods. No “fully” direct methods exist in forestry. Basically, the distinction is between those based on direct measurement, or contact-approaches (Jonckheere et al. 2004) and those based on optical theory. With “Indirect methods “ the literature consider all the methods based on optical theory, and particularly the Beer-Lambert law. Indirect methods in the field are based on both passive and active sensors.
Importantly, authors should explain the difference between transmittance based and gap fraction based methods. The former requires measuring above and below (absolute or relative) light, and then calculate light transmittance accordingly. This comprises PAR ceptometers . Gap fraction based methods include LAI-2000, which estimate gap as the ratio of above to below canopy relative radiation, and DHP, which calculate gap fraction as the ratio of classified pixels gaps over total image pixels. With some assumption (e.g., leaves are homogeneous turbid medium) gap fraction and light transmittance equals, and then the Beer Lambert law is used on all these methods.
Finally, I would add in the intro the well-established Beer-law formula and explain differences between effective to true LAI in optical measurements. This will help introducing the importance of clumping index, its multi-scale spatial nature, and why different methods provide different clumping corrections, and finally, why effective LAI is the first step to know the true LAI, also compared with those from direct methods (which by definition ignore clumping)
L33 the litter traps is also labor intensive cause the litter should be separated, and the leaf component dried in forced-air stove, to determine the dry mass, which is combined with SLA.
L 60 here an example of continuous DHP measurements https://www.sciencedirect.com/science/article/pii/S0168192320300460
L90. Timestemp/timestep -> sampling period
L90-120 From the best I can understand this approach foresee the indirect (i.e. scanning) measurements of all leaves in the traps. It is noticeably more time consuming than measuring a sample of leaves to determine SLA and then multiply with the total biomass. Authors should acknowledge this. Additionally, it is the very first time I see a litter trap laying at the floor - these baskets are typically set at 1 m height to reduce seed predation - how these traps be impacted by fauna, which can influence the leaf litter inside?
Citation: https://doi.org/10.5194/egusphere-2025-1496-RC1 -
AC2: 'Reply on RC1', Simon Lotz, 09 Jul 2025
reply
Dear Francesco Chianucci,
Thank you for the constructive and detailed feedback. We particularly appreciate your comments on (1) the influence of image acquisition conditions—such as sky exposure, automatic settings, and gamma correction—on DHP-derived LAI estimates, and (2) the need to explore how gap fraction, clumping index, and G(θ) vary with zenith angle to better understand LAI inversion. These important points as well as the specific comments will be carefully considered and incorporated into the revised manuscript.
Citation: https://doi.org/10.5194/egusphere-2025-1496-AC2
-
AC2: 'Reply on RC1', Simon Lotz, 09 Jul 2025
reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 461 | 87 | 17 | 565 | 14 | 29 |
- HTML: 461
- PDF: 87
- XML: 17
- Total: 565
- BibTeX: 14
- EndNote: 29
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper made a simple exercise to compare LAI obtained from digital hemispherical photography (DHP) and litter traps (LT) and proposed a linear model to adjust the original DHP measurements. I found the contribution limited for consideration.
GENERAL COMMENTS
The innovation of the manuscript is limited. This work closely follows another previous study by Liu et al. 2015b (doi:10.1139/cjfr-2014-0351). However, the previous paper simply performed empirical woody and clumping correction for deciduous broadleaf forest in order to match DHP LAI with litter trap observations. Such cite-specific adjustment is not generic.
The paper states that “it remains unclear if the DHP method enables to robustly track temporal LAI dynamics” (L4-5). Authors need to get familiar with current progress of using DHP for temporal LAI measurement. There are many related studies such as
(doi: 10.1016/j.agrformet.2014.08.005, doi: 10.1016/j.agrformet.2018.02.003) for seasonal crop LAI measurement with DHP. There are even many automatic DHP measurement studies:
https://doi.org/10.1016/j.agrformet.2022.108999
https://doi.org/10.1016/j.agrformet.2020.107944
https://doi.org/10.1111/2041-210X.14199
SPECIFIC COMMENTS
L19 Note that the “total intercepting area” is different from the flat area (L93). LAI is defined for the flat area, not intercepting area.
L100 For “the cumulative LAI”, do you mean “the cumulative LT LAI”?
L130-138. The LXG method is essentially different from the LX method is the estimation of clumping index (Fang, 2021; doi: 10.1016/j.agrformet.2021.108374). The LXG CI is not a ratio of effective LAI to the true LAI.
Section 2.5
The generalized linear mixed model (GLMM) was not clearly introduced. There are intermediate steps not fully presented. What and how are the inverse Gaussian distribution and a log link function applied?
L187. For Fig. 5 here, do you mean Fig. 4?
Section 3.1
The comparison of DHP and LT LAI was not clearly presented. It’s recommended to show a scatterplot to compare both DHP and LT LAI observations. Also show the effective LAI scatterplot and the clumping index derived from different view zenith angles.
L191-194 can be moved to section 3.1. I guess the Fig. 3 in L193 should be read as Fig. 4.
Section 3.3
I would suggest to show the slopes and intercepts (Eq. (1)) for different phases.
Section 4.1
I would not use the term “spatial footprint of LTs” since footprint is mostly used for LiDAR observation in this community. LT data are supposed to represent the whole sample plot.