the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 May 2025
Biases in estimated vegetation indices from observations under cloudy conditions
Abstract. Field observations of vegetation indices (VIs) from drones and aircraft provide higher spatial resolution than satellites. Vegetation indices are derived from ratios of spectral reflectivity measurements. The reflectivity is measured in a relative way by periodic reference measurements over reflectance panels. This requires cloud-free or at least stable cloud conditions between reflectance panel measurements. This assumption is often violated, with the effect that wavelength-dependent scattering and absorption of radiation by clouds lead to a distortion of the below-cloud spectral downward irradiance F↓(λ) and thus affects estimates of VIs.
This paper presents combined atmosphere-vegetation radiative transfer (RT) simulations to systematically investigate cloud-induced biases in remotely sensed VIs derived from below-cloud measurements. The biases in VIs have been investigated for the general case of two-band VIs, and for the special cases of the normalized difference vegetation index (NDVI), the normalized difference water index (NDWI), and the enhanced vegetation index (EVI). For the general case of two-band VIs the lowest sensitivity to cloud changes was found for wavelength combinations below 1400 nm and outside the water vapor absorption bands. The NDVI was found to be almost insensitive to changes in cloud conditions, while greater biases were identified for the NDWI. The EVI was also found to be susceptible to cloud changes, leading to biases of 0.36 in the selected example with biases in the estimated leaf area index of 1.3.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1834 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1834 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-2082', Anonymous Referee #1, 06 Jul 2025
- AC1: 'Reply on RC1', Kevin Wolf, 30 Jul 2025
-
RC2: 'Comment on egusphere-2025-2082', Anonymous Referee #2, 14 Jul 2025
This manuscript reports on the effect of clouds on the derivation of vegetation indices from remote sensing platforms in a radiative transfer modelling approach. Basically it is a sensitivity study on how clouds contribute to the incoming radiation and what effect it might have on the calibration conditions for deriving the reflectance values and consequently on the computation of the various VI with Sentinel 2 bands as example for different solar zenith angles.
Overall, it is an interesting and relevant study since it brings our attention to the basics of using reflectance values and its impact on derived vegetation indices. It is important that we do not forget the basics of RS data. The preprint is very well written and well-documented, with nice figures.
To my opinion, the authors should elaborate a bit more on the consequences of the effect on biases in values of VI’s. They briefly touch the impact on LAI derived from EVI, but the authors should add some paragraph how it could impact remotely sensed derived biophysical properties such as GPP, green water fluxes etc with proper referencing. This would make the paper even more relevant. Perhaps, indicating some of the effects as percentages can increase the visibility.
In lines 117, 306 and others, the authors directly link NDVI to vegetation health. I would refrain from that, since NDVI essentially says something about the “greenness” of the vegetation, but not necessarily on its health. It can be used for vegetation health, but it is not synonymous for health.
Furthered, I only have few minor/textual comments/
- Normally, numbers less than 10 are written as text; So less than one; equals one, etc
- If one uses for instance 0.29, also use 0.20 and not 0.2 (see L338, but also elsewhere): always use the same amount of decimals for the same property;
- I do not know the policy of the journal, but normally the cited references in the text are first ordered chronologically and then alphabetically;
- Abstract: add more on possible consequences;
- Captions Fig 1. Replace “‘relate” with “connect”;
- L30: specify what tau is under the text of Eq 1; see lines 45-47; this should come earlier in the text;
- L61: “attempts”? This is not a proper use of the word here;
- L67: In general, it is a “transfer function” rather than a factor, although used as a factor;
- L76: Should be “requires frequent calibrations of the transfer function”;
- L217: remove “the“ before Appendix A;
- L224: “valueS”;
- L270-271: “ARE close to zero”;
Figure 5: Why is the symbol of NDII missing in the upper right panel?
- L281, 324 and other lines: I would refrain of using the term “exemplary”; it echoes a bit as “exemplary behaviour or punishment”; Perhaps use “illustrates”?
- L345: Should be “The effect of changes in fdir”;
- L378-379: “… was between 0 and 40,” and “.. ranged …”; Remove “were covered”:
- L406: remove comma after NDWI
- L407: twice increase, increasing;
Citation: https://doi.org/10.5194/egusphere-2025-2082-RC2 - AC2: 'Reply on RC2', Kevin Wolf, 30 Jul 2025
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-2082', Anonymous Referee #1, 06 Jul 2025
Dear Dr. Wolf and co-authors,
Thank you for developing this framework to address the cloud influence on ground reflectance measurements. Indeed, in the remote sensing community, our field measurements are highly dependent on the illumination conditions, frequently altered by clouds. I have several minor remarks for your consideration.
In eq. 1, 4, 6 what do you mean by the ‘sr’ argument? Pi is already assumed to be in steradian (sr) units, cancelling steradian in the upwelling radiance I.
A bit on the same line, multiplication by pi suggests that the surface reflects homogeneously in all directions, Lambertian reflectance. How big would you expect the influence of the directionality of the actual surface to be on the reflectance value?
On the SCOPE model (section 2.2.2 and Table 2).
- First of all, it is unclear why SCOPE was chosen instead of SAIL or INFORM. The latter is more suitable for pine forest (L163) simulations, as it explicitly has the concept of trunks and branches in it. SCOPE has energy balance, thermal domain, photosynthesis and chlorophyll fluorescence that other models do not have. Using it as a tool for a single reflectance simulation is overkill. Nonetheless, your choice.
- Why was only half of the important SCOPE input parameters chosen? The BSM model, for example, has brightness, two shape parameters and moisture content, but only B is shown in Table 2. In any case, those parameters are set to their default values so it is also not clear why highlighting them at all.
- Finally, could you please be more explicit about which SCOPE output was integrated with the libRadtran output and how? I am a bit confused because L152 says “As an initial guess of the surface albedo in libRadtran, the "mixed-forest" albedo was taken from the IGBP data base.” According to my understanding, it was sufficient to take some spectral forest reflectance for the exercise instead of running an RTM.
Figure 2b. Please, add a legend.
L187 – “Wolf et al. (2024) have shown the influence of clouds on direct and diffuse F↓(λ), the associated effects on F↑(λ),” please, write exactly what the influence was. I guess more clouds – more diffuse radiation.
Figure 3. What do grey areas show? Sentinel-2 bands?
Figures 3 and 4 captions. Please, note, you are working with synthetic (modelled) data. Remove the term “measured” reflectance; do not mislead the readers.
Figure 5. Please, check the location of symbols inside the heatmaps Whereas for NDVI (circle) lambda1 and lambda2 are matching the expected NIR and RED, NDWI1240 is definitely far from lambda1=1240 nm. Furthermore, the symbol in Figures 5c and 5d around lambda1=900nm, lambda2=1600nm is unclear (or absent from the legend).
Citation: https://doi.org/10.5194/egusphere-2025-2082-RC1 - AC1: 'Reply on RC1', Kevin Wolf, 30 Jul 2025
-
RC2: 'Comment on egusphere-2025-2082', Anonymous Referee #2, 14 Jul 2025
This manuscript reports on the effect of clouds on the derivation of vegetation indices from remote sensing platforms in a radiative transfer modelling approach. Basically it is a sensitivity study on how clouds contribute to the incoming radiation and what effect it might have on the calibration conditions for deriving the reflectance values and consequently on the computation of the various VI with Sentinel 2 bands as example for different solar zenith angles.
Overall, it is an interesting and relevant study since it brings our attention to the basics of using reflectance values and its impact on derived vegetation indices. It is important that we do not forget the basics of RS data. The preprint is very well written and well-documented, with nice figures.
To my opinion, the authors should elaborate a bit more on the consequences of the effect on biases in values of VI’s. They briefly touch the impact on LAI derived from EVI, but the authors should add some paragraph how it could impact remotely sensed derived biophysical properties such as GPP, green water fluxes etc with proper referencing. This would make the paper even more relevant. Perhaps, indicating some of the effects as percentages can increase the visibility.
In lines 117, 306 and others, the authors directly link NDVI to vegetation health. I would refrain from that, since NDVI essentially says something about the “greenness” of the vegetation, but not necessarily on its health. It can be used for vegetation health, but it is not synonymous for health.
Furthered, I only have few minor/textual comments/
- Normally, numbers less than 10 are written as text; So less than one; equals one, etc
- If one uses for instance 0.29, also use 0.20 and not 0.2 (see L338, but also elsewhere): always use the same amount of decimals for the same property;
- I do not know the policy of the journal, but normally the cited references in the text are first ordered chronologically and then alphabetically;
- Abstract: add more on possible consequences;
- Captions Fig 1. Replace “‘relate” with “connect”;
- L30: specify what tau is under the text of Eq 1; see lines 45-47; this should come earlier in the text;
- L61: “attempts”? This is not a proper use of the word here;
- L67: In general, it is a “transfer function” rather than a factor, although used as a factor;
- L76: Should be “requires frequent calibrations of the transfer function”;
- L217: remove “the“ before Appendix A;
- L224: “valueS”;
- L270-271: “ARE close to zero”;
Figure 5: Why is the symbol of NDII missing in the upper right panel?
- L281, 324 and other lines: I would refrain of using the term “exemplary”; it echoes a bit as “exemplary behaviour or punishment”; Perhaps use “illustrates”?
- L345: Should be “The effect of changes in fdir”;
- L378-379: “… was between 0 and 40,” and “.. ranged …”; Remove “were covered”:
- L406: remove comma after NDWI
- L407: twice increase, increasing;
Citation: https://doi.org/10.5194/egusphere-2025-2082-RC2 - AC2: 'Reply on RC2', Kevin Wolf, 30 Jul 2025
Peer review completion
Journal article(s) based on this preprint
Data sets
Simulated spectral irradiances, radiances, and vegetation albedo obtained from coupling libRadtran and SCOPE2.0 Kevin Wolf et al. https://doi.org/10.5281/zenodo.15275610
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 775 | 101 | 26 | 902 | 24 | 35 |
- HTML: 775
- PDF: 101
- XML: 26
- Total: 902
- BibTeX: 24
- EndNote: 35
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Evelyn Jäkel
André Ehrlich
Michael Schäfer
Hannes Feilhauer
Andreas Huth
Manfred Wendisch
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1834 KB) - Metadata XML
Dear Dr. Wolf and co-authors,
Thank you for developing this framework to address the cloud influence on ground reflectance measurements. Indeed, in the remote sensing community, our field measurements are highly dependent on the illumination conditions, frequently altered by clouds. I have several minor remarks for your consideration.
In eq. 1, 4, 6 what do you mean by the ‘sr’ argument? Pi is already assumed to be in steradian (sr) units, cancelling steradian in the upwelling radiance I.
A bit on the same line, multiplication by pi suggests that the surface reflects homogeneously in all directions, Lambertian reflectance. How big would you expect the influence of the directionality of the actual surface to be on the reflectance value?
On the SCOPE model (section 2.2.2 and Table 2).
Figure 2b. Please, add a legend.
L187 – “Wolf et al. (2024) have shown the influence of clouds on direct and diffuse F↓(λ), the associated effects on F↑(λ),” please, write exactly what the influence was. I guess more clouds – more diffuse radiation.
Figure 3. What do grey areas show? Sentinel-2 bands?
Figures 3 and 4 captions. Please, note, you are working with synthetic (modelled) data. Remove the term “measured” reflectance; do not mislead the readers.
Figure 5. Please, check the location of symbols inside the heatmaps Whereas for NDVI (circle) lambda1 and lambda2 are matching the expected NIR and RED, NDWI1240 is definitely far from lambda1=1240 nm. Furthermore, the symbol in Figures 5c and 5d around lambda1=900nm, lambda2=1600nm is unclear (or absent from the legend).