the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Feb 2023
Hyperspectral imaging sediment core scanning tracks high-resolution Holocene variations in (an)oxygenic phototrophic communities at Lake Cadagno, Swiss Alps
Abstract. Pigments produced by anoxygenic phototrophic bacteria are valuable proxies of past anoxia in lacustrine and marine environments. Pigment measurement typically requires time-consuming and costly chemical extractions and chromatographic analyses, which limits the temporal resolution of paleoenvironmental reconstructions based on sedimentary pigments. Here, we evaluate the potential of in-situ hyperspectral imaging (HSI) core scanning as a rapid, non-destructive method to document high-resolution changes in oxygenic and anoxygenic phototrophic communities at meromictic Lake Cadagno, Switzerland. Three distinct groups of pigments can be detected with the HSI method in the sediments of Lake Cadagno; each pigment group represents a different phototrophic community. Oxygenic phototrophs are indicated by total chloropigments (TChl; chlorophyll-a, -b and derivatives). Two types of anoxygenic phototrophs were distinguished – purple sulfur bacteria (PSB), represented by bacteriochlorophyll-a, and green sulfur bacteria (GSB), represented by bacteriochlorophylls-c, -d, and -e. HSI pigment indices were validated by pigment measurements performed on extracted samples using spectrophotometer and high-performance liquid chromatography (HPLC). Bacteriochlorophylls were present throughout the past 10 kyr, confirming geochemical evidence of nearly continuous stratification and sulfidic conditions at Lake Cadagno. Major shifts in the anoxygenic phototropic communities are recorded at decadal to millennial scales. GSB and PSB communities coexisted from 10.2–3.4 kyr BP, with dominance of PSB over GSB from 8.8–3.4 kyr BP indicating strongly stratified conditions in the lake and strong light radiation at the chemocline. From 3.4–1.3 kyr BP, PSB were mostly absent, and GSB became dominant, implying lower light intensity at the chemocline due to a combination of factors including deforestation in the lake surroundings, increased flood frequency, cooler climatic conditions, and changes in groundwater solute concentrations. The high-resolution HSI data show that frequent flood events and mass movements disturbed the chemocline and the anoxygenic bacterial communities, and that the PSB were particularly sensitive and slow to recover following these disturbance events. This study demonstrates for the first time that HSI can detect GSB related pigments, making the method uniquely valuable as a rapid tool to study samples containing pigments of both oxygenic and anoxygenic phototrophs.
-
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
(1700 KB)
-
Supplement
(280 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1700 KB) - Metadata XML
-
Supplement
(280 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-151', Anonymous Referee #1, 09 Mar 2023
The paper is well-written and the idea is ueful for the community. however, there are some doubts about reliability of the new proposed index which require more debates statistically. Here are my comments:
RABD670: There is a review from Van Exem et al. 2022 which they reviewed all the chl indices and they found out that RABAs work better than RABDs. Have you tried to implement that and compare the results since it seems to be possible with your data.
Line 160: how did you determine the percentage reflectance values to remove cracks and mineral reflections? it seems it is based on practical excercise on your core and your hyperspectral scanner which it should be noted in the text.
Fig 2: since RABD710-730 is located in the right shoulder of chl-a and phaephytin-a absorption bands, how would they effect this index? with bigger amount of chl-a and Phae-a it is expected that GSB-related signal disapear. And based on your data was there any detection limitation on this index?
Fig3: It would be better to remake all the plots in a way that the whole dots and the range are observed.
Fig 3C: There are some RABD710-730 dots which are valued less than one, how do you interpret them? moreover, considering RABD710-730, I am curios to see what correlation you will get if you remove Isorenieratene which have big values ( over 1000) and then recalculate the correaltion. It seems the correlation for the samples with Iso under 1000 are weak. Maybe, it can be discussed in terms of limitations of the index!Fig4: this figure is a bit unclear. suggest to change the caption and specify y-axes is related to which RABD.
Finally, it is always a question that a model or here an index which is applied on one core can be applied on any other cores? what would be the limitations and maybe a discussion on this in the paper would be useful for reader.
Citation: https://doi.org/10.5194/egusphere-2023-151-RC1 -
AC1: 'Reply on RC1', Paul Zander, 28 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-151/egusphere-2023-151-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Paul Zander, 28 Apr 2023
-
RC2: 'Comment on egusphere-2023-151', Anonymous Referee #2, 10 Apr 2023
Overview
Zander et al. use recently developed hyperspectral imaging methods, including one first introduced in this manuscript, to quantify the abundance of sedimentary photopigment produced by oxygenic phototrophs and anoxygenic phototrophs - specifically purple and green sulfur bacteria (PSB and GSB, respectively). They apply this technique to Lake Cadagno, a lake that has been well studied for both modern and paleolimnology. The authors find that total chloropigments (TChl) and purple sulfur bacteria (PSB) are readily detectable using key absorption troughs in hyperspectal imaging, and calibrate their abundance using concentrations measured on samples using extraction and spectrophotometry and HPLC. They argue that GSB are detectable by a characteristic absorption trough, and around 700 nm, however it is difficult to calibrate as the pigments are not readily measured by other techniques. Instead they use co produced carotenoids to validate the detection of GSB, but do not attempt a calibration, instead they interpret the index as an indicator of relative abundance.
The detection and interpretation of these pigments by hyperspectral imaging has great potential for paleolimnological studies. HSI methods are rapid, high-resolution and non destructive, and analyzing these pigments provides important and direct inference on the past light and oxygenation structure of the lake. GSB is also quite difficult to measure, and it’s possible that in situ detection via HSI may actually work better than other approaches, although this is difficult to prove. In the case of Lake Cadagno, the pigment records provide important insights on both long-term lake evolution, and response to external forcings and pressures, but also the response of the lake to flood events, highlighting the value of the high resolution detection.
The manuscript is well written and well illustrated, and effectively articulates the importance and value of the approach. It’s an important contribution to the paleolimnologic literature, and I recommend it for publication following the correction of a minor concerns.
Minor concerns:
Non-normal distributions of pigment and RABD data
My only substantial concern with analysis and the results has to do with the treatment of the data in the calibration and reconstruction approaches. Specifically, I worry about non-normality in both the the pigment concentrations measured by spectroscopy and HPLC, and in the RABD indices. Both datasets are left bounded, and tend to be skewed right. Some strongly so. On line 236 the authors note that this may hamper their interpretations. The authors use linear regression to relate RABD indices to TChla and PSB, and to quantify their uncertainties, but this approach assumes normality in both the predictor and predictand. Looking at figure 3, this seems reasonable for TChl-a and RABD670, but the PSB data look right skewed, and potentially bimodal. And looking at figure 5, all the downcore HSI data, and especially the GSB index and PSB reconstruction both look left bounded and right skewed, as expected. I’d encourage the authors to either demonstrate normality for these data, or consider transforming them before calibration, or use a different approach that does not require normally-distributed data.
For the Bphe-a and RABD842, I wonder whether the data really support a continuous calibration, or whether a on/off, presence/absence type calibration would be more appropriate. To my eye (Fig 3A), there are basically two populations of Bphe-a concentrations. Those close to zero, and those around 1000 ug/g. The lower concentration population does suggest a relationship, but the trend is notably lower than that inferred from the dataset as a whole. The two highest Bphe-a concentrations do also stand out, but it’s hard to know how much to weight those outliers. I’m open to the argument for a continuous relationship, but I think it needs to be made as it seems plausible that we’re mostly seeing presence/absence of PSB in these data.
Other Concerns
Figure 4.
The comparison between the aeDNA data and the PSB and GSB indices is very interesting. I like the presentation in figure 4, but would be very keen to see a direct comparison of the aeDNA sample data and the HSI-based ratio. Could you add a scatter plot where the PSB/GSB indices over the DNA sample depth ranges are averaged and the plotted against a similar metric in the DNA data? There are intervals where it seems to agree quite well, but others where it doesn’t, and I’d like to see the comparison explicitly, while also accounting for the different resolution of the datasets.
Figure 5. Add a Tchl label to panel A to match panels B and C.
Figure 6. Continuing the Tchl, PSB and GSB labeling on the curves would benefit figure 6 too.
Citation: https://doi.org/10.5194/egusphere-2023-151-RC2 -
AC2: 'Reply on RC2', Paul Zander, 28 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-151/egusphere-2023-151-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Paul Zander, 28 Apr 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-151', Anonymous Referee #1, 09 Mar 2023
The paper is well-written and the idea is ueful for the community. however, there are some doubts about reliability of the new proposed index which require more debates statistically. Here are my comments:
RABD670: There is a review from Van Exem et al. 2022 which they reviewed all the chl indices and they found out that RABAs work better than RABDs. Have you tried to implement that and compare the results since it seems to be possible with your data.
Line 160: how did you determine the percentage reflectance values to remove cracks and mineral reflections? it seems it is based on practical excercise on your core and your hyperspectral scanner which it should be noted in the text.
Fig 2: since RABD710-730 is located in the right shoulder of chl-a and phaephytin-a absorption bands, how would they effect this index? with bigger amount of chl-a and Phae-a it is expected that GSB-related signal disapear. And based on your data was there any detection limitation on this index?
Fig3: It would be better to remake all the plots in a way that the whole dots and the range are observed.
Fig 3C: There are some RABD710-730 dots which are valued less than one, how do you interpret them? moreover, considering RABD710-730, I am curios to see what correlation you will get if you remove Isorenieratene which have big values ( over 1000) and then recalculate the correaltion. It seems the correlation for the samples with Iso under 1000 are weak. Maybe, it can be discussed in terms of limitations of the index!Fig4: this figure is a bit unclear. suggest to change the caption and specify y-axes is related to which RABD.
Finally, it is always a question that a model or here an index which is applied on one core can be applied on any other cores? what would be the limitations and maybe a discussion on this in the paper would be useful for reader.
Citation: https://doi.org/10.5194/egusphere-2023-151-RC1 -
AC1: 'Reply on RC1', Paul Zander, 28 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-151/egusphere-2023-151-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Paul Zander, 28 Apr 2023
-
RC2: 'Comment on egusphere-2023-151', Anonymous Referee #2, 10 Apr 2023
Overview
Zander et al. use recently developed hyperspectral imaging methods, including one first introduced in this manuscript, to quantify the abundance of sedimentary photopigment produced by oxygenic phototrophs and anoxygenic phototrophs - specifically purple and green sulfur bacteria (PSB and GSB, respectively). They apply this technique to Lake Cadagno, a lake that has been well studied for both modern and paleolimnology. The authors find that total chloropigments (TChl) and purple sulfur bacteria (PSB) are readily detectable using key absorption troughs in hyperspectal imaging, and calibrate their abundance using concentrations measured on samples using extraction and spectrophotometry and HPLC. They argue that GSB are detectable by a characteristic absorption trough, and around 700 nm, however it is difficult to calibrate as the pigments are not readily measured by other techniques. Instead they use co produced carotenoids to validate the detection of GSB, but do not attempt a calibration, instead they interpret the index as an indicator of relative abundance.
The detection and interpretation of these pigments by hyperspectral imaging has great potential for paleolimnological studies. HSI methods are rapid, high-resolution and non destructive, and analyzing these pigments provides important and direct inference on the past light and oxygenation structure of the lake. GSB is also quite difficult to measure, and it’s possible that in situ detection via HSI may actually work better than other approaches, although this is difficult to prove. In the case of Lake Cadagno, the pigment records provide important insights on both long-term lake evolution, and response to external forcings and pressures, but also the response of the lake to flood events, highlighting the value of the high resolution detection.
The manuscript is well written and well illustrated, and effectively articulates the importance and value of the approach. It’s an important contribution to the paleolimnologic literature, and I recommend it for publication following the correction of a minor concerns.
Minor concerns:
Non-normal distributions of pigment and RABD data
My only substantial concern with analysis and the results has to do with the treatment of the data in the calibration and reconstruction approaches. Specifically, I worry about non-normality in both the the pigment concentrations measured by spectroscopy and HPLC, and in the RABD indices. Both datasets are left bounded, and tend to be skewed right. Some strongly so. On line 236 the authors note that this may hamper their interpretations. The authors use linear regression to relate RABD indices to TChla and PSB, and to quantify their uncertainties, but this approach assumes normality in both the predictor and predictand. Looking at figure 3, this seems reasonable for TChl-a and RABD670, but the PSB data look right skewed, and potentially bimodal. And looking at figure 5, all the downcore HSI data, and especially the GSB index and PSB reconstruction both look left bounded and right skewed, as expected. I’d encourage the authors to either demonstrate normality for these data, or consider transforming them before calibration, or use a different approach that does not require normally-distributed data.
For the Bphe-a and RABD842, I wonder whether the data really support a continuous calibration, or whether a on/off, presence/absence type calibration would be more appropriate. To my eye (Fig 3A), there are basically two populations of Bphe-a concentrations. Those close to zero, and those around 1000 ug/g. The lower concentration population does suggest a relationship, but the trend is notably lower than that inferred from the dataset as a whole. The two highest Bphe-a concentrations do also stand out, but it’s hard to know how much to weight those outliers. I’m open to the argument for a continuous relationship, but I think it needs to be made as it seems plausible that we’re mostly seeing presence/absence of PSB in these data.
Other Concerns
Figure 4.
The comparison between the aeDNA data and the PSB and GSB indices is very interesting. I like the presentation in figure 4, but would be very keen to see a direct comparison of the aeDNA sample data and the HSI-based ratio. Could you add a scatter plot where the PSB/GSB indices over the DNA sample depth ranges are averaged and the plotted against a similar metric in the DNA data? There are intervals where it seems to agree quite well, but others where it doesn’t, and I’d like to see the comparison explicitly, while also accounting for the different resolution of the datasets.
Figure 5. Add a Tchl label to panel A to match panels B and C.
Figure 6. Continuing the Tchl, PSB and GSB labeling on the curves would benefit figure 6 too.
Citation: https://doi.org/10.5194/egusphere-2023-151-RC2 -
AC2: 'Reply on RC2', Paul Zander, 28 Apr 2023
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-151/egusphere-2023-151-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Paul Zander, 28 Apr 2023
Peer review completion
Journal article(s) based on this preprint
Data sets
Lake Cadagno sediment core hyperspectral imaging and pigment data tables Zander, P. D., Wirth, S. B., Gilli, A., Peduzzi, S., and Grosjean, M. https://doi.org/10.5281/zenodo.7573508
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 374 | 100 | 17 | 491 | 34 | 2 | 5 |
- HTML: 374
- PDF: 100
- XML: 17
- Total: 491
- Supplement: 34
- BibTeX: 2
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Stefanie B. Wirth
Adrian Gilli
Sandro Peduzzi
Martin Grosjean
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1700 KB) - Metadata XML
-
Supplement
(280 KB) - BibTeX
- EndNote
- Final revised paper