A 900-Year Isotopic Proxy Rainfall Record from Northeastern Botswana
, Grant Hall 3, Christiaan W. Winterbach 4, Iain Robertson 5, Ileana Andreea Ratiu 1,2, Victor Bocos-Bintintan 6, Laszlo Rakosy 7 and Stephan Woodborne 8Round 1
Reviewer 1 Report
The presented manuscript is interesting but there are several, importa weak points which require a major revision before considering it for publication.
Here are some detailed comments.
a. Why in Table 1 some of the results are presented with 1 sigma and other with 2 sigma confidence level? Please choose one. Typically the 2sigma range should be considered.
b. It is not clear what the "assigned year" is. How its was calculated? Is it a median value? In any case I do not understand how it is possible that no uncertainty is included. Furthermore calibrated time ranges are typically non Gaussian and Multimodal, and the complexity of interpretation of these data cannot be avoided by simply masking the uncertainty.
c. I have also concerns about the used age model. First it is not very clear how this model is implemented and on which hypotheses it is based. The main idea seems to be that a linear correlation exist between depth into the tree and age. This assumes a constant radial growth rate. Is this correct? How strong this hypothesis is. Is it supported by literature? This should be well discussed. In any case the application of a model does not cancel by itself the uncertainty of a measurement! There are several, statistically robust modeling tools which can be used to properly estimate the uncertainties such as those based on Bayesian modeling. For instance by using the OxCal software.
d. Radiocarbon dates were calibrated by using the last internationally accepted atmospheric curve? Is this correct? Please cite.
e. Please add a reference to a review paper about radiocarbon. For instance Hajdas, I., Ascough, P., Garnett, M.H. et al. Radiocarbon dating. Nat Rev Methods Primers1, 62 (2021). https://doi.org/10.1038/s43586-021-00058-7
f. How many samples were analyzed for d13C?
g. It is not very clear how Fig. 2 was obtained. In particular how the age of each sample submitted to d13C analyses was estimated? If it was done by using the age model described before then the comment about uncertainty should be considered.
g. The discussion about the correlation between the measured isotopic data and the literature rainfall records seems to me quite weak. In particular considering that only agreement index (R) is reported but to an in depht analysis is missing. For instance by putting the data on the same graph.
Author Response
Point-by-point response to Reviewer 1
General remark: The presented manuscript is interesting but there are several, importa weak points which require a major revision before considering it for publication.
Response to general remark: Thank you for evaluating and helping to consolidate our manuscript
Remark a: Here are some detailed comments.
- Why in Table 1 some of the results are presented with 1 sigma and other with 2 sigma confidence level? Please choose one. Typically the 2sigma range should be considered.
Response to remark a: The radiocarbon dates and the corresponding errors were provided in the table. The confidence interval for obtaining calibrated age ranges is not subject to any international reporting rules. However, we adjusted the age ranges, which are now presented using only the 2-sigma confidence level.
Remark b: It is not clear what the "assigned year" is. How its was calculated? Is it a median value? In any case I do not understand how it is possible that no uncertainty is included. Furthermore calibrated time ranges are typically non Gaussian and Multimodal, and the complexity of interpretation of these data cannot be avoided by simply masking the uncertainty.
Response to remark b:
We have rewritten subsection 2.5. concerning the age model of the Chapman baobab as follows:
“In the absence of clear annual rings, an age model for the δ13C time series was constructed from the radiocarbon analyses. We make the assumption that two baobab stems from the same tree experienced the same environmental variation through time and should therefore have matching δ13C time series. The match between the two δ13C time series is therefore used as a constraint in the formulation of the age model. Unfortunately, the available Bayesian age/depth models that use the known age relationship between samples (whether one radiocarbon sample is older or younger than another is known a priori) are not able to accommodate the match between δ13C time series as a priori constraints. Instead we use a series of linear age relationship between aliquot number and date (that is, assuming a linear relationship between radial growth and age) that meets the constraint that the model for each individual stem should intercept the radiocarbon ages within the 2-sigma error range, and that excursions in the δ13C time series between stems should match. Assuming that the radial growth rate is constant over extended periods of time is obviously flawed, as the trees likely grow faster under favourable conditions and slower under unfavourable conditions, and this means that the age model only approximates the average growth over time. This error is constrained by the 2-sigma match of the age model with the radiocarbon dates, but the age model will have an inherent under- and over-approximation of the growth rate that introduces errors in the age assignment for individual aliquots in the δ13C time series. For this reason we follow the method used by [20,64,65] and apply a 21-year biweight mean (a mean δ13C value that is weighted by the δ13C values from both stems for the 10 years prior to, and after, the aliquot in question) [80] The biweight mean also has an associated variance calculation, but this is substantially smaller than the errors in the radiocarbon dates, and while we consider it, it is not presented in the result.
This approach allows a date to be assigned to each aliquot in δ13C time series, but it is necessary to caution against a literal interpretation of the dates. The error range in radiocarbon dating limits precision to approximately 30 years and this must be combined with the failure of the age model to approximate minor growth rate variations. The errors are less problematic with longer time series where the variability is assessed at multi-decadal to centennial ranges, but it does have a strong effect on correlating the younger part of the record with instrumental data. Even though there is an acknowledged weakness in the age model, the date errors on adjacent aliquots will be similar, and the resulting time series of water stress in the trees still provides a strong indication of the frequency and duration of protracted droughts and wet periods.
When the age model has been applied to the δ13C time series it is necessary to make a correction for the temporal changes in the atmospheric δ13C variation through time. We used the Southern Hemisphere record of atmospheric δ13C variability [80] with online updates of the dataset (http://cdiac.ornl.gov/ftp/db1014/isotope.cgo). Changes in the intrinsic water use efficiency (iWUE) of trees has occurred in response to elevated CO2 concentrations post the industrial age [81], but Woodborne et al. [20] proved that the overall effect of such iWUE changes have a relatively small effect on the isotope proxy record, and are irrelevant in the calculation.”
Remark c: I have also concerns about the used age model. First it is not very clear how this model is implemented and on which hypotheses it is based. The main idea seems to be that a linear correlation exist between depth into the tree and age. This assumes a constant radial growth rate. Is this correct? How strong this hypothesis is. Is it supported by literature? This should be well discussed. In any case the application of a model does not cancel by itself the uncertainty of a measurement! There are several, statistically robust modeling tools which can be used to properly estimate the uncertainties such as those based on Bayesian modeling. For instance by using the OxCal software.
Response to remark c: The only literature on baobab radial growth is derived from our extensive project on baobab dating. They have ring structures but out analyses have shown that rings are seldom annually formed. There are several published records using linear radial growth model. Again this is our research but it has passed review in several highly ranked journals, including in Radiocarbon. In every case where we have dated baobabs, the radial growth is linear within the generous constraints of the radiocarbon method. We acknowledge this, but feel that the emphasis of the manuscript is not to assess sub-decadal variability in the climate record, but rather to focus on the longer term trends where the acknowledged shortcomings of the age model are less relevant.
The use of Bayesian age models is difficult, and the text has been expanded to elaborate on this (see our previous response to remark b).
Remark d. Radiocarbon dates were calibrated by using the last internationally accepted atmospheric curve? Is this correct? Please cite.
Response to remark d: This information is stated in the last paragraph of subsection 2.3. Radiocarbon dating: Pretreatment, AMS measurements and calibration: “The obtained fraction modern values were converted to a radiocarbon date. Radiocarbon dates and errors were rounded to the nearest year. Radiocarbon dates were calibrated and converted into calendar dates with the OxCal v4.4 program for Windows [77] and the Southern Hemisphere terrestrial dataset SHCal20 [78].”
Remark e. Please add a reference to a review paper about radiocarbon. For instance Hajdas, I., Ascough, P., Garnett, M.H. et al. Radiocarbon dating. Nat Rev Methods Primers 1, 62 (2021). https://doi.org/10.1038/s43586-021-00058-7
Response to remark e: Thank you for your suggestion. This reference was introduced in the text as reference 56.
Remark f. How many samples were analyzed for d13C?
Response to remark f: We have changed the last paragraph of subsection 2.4 referring to the stable isotope analysis to express more clearly that 2236 δ13C measurements were performed:
“Sub-samples of each aliquot (0.05-0.06 x 10-6 kg) were placed in tin cups for the isotopic analysis. The δ13C measurements were performed for all 2236 aliquot subsamples at the University of Pretoria, South Africa, using a Thermoquest EA 1110 elemental analyser coupled with a Delta V Plus isotope ratio mass spectrometer with a ConFlo III interface. Sub-samples of each aliquot (0.05-0.06 x 10-6 kg) were placed in tin cups for the isotopic analysis. Laboratory standard samples (Shorea superba) and blanks were run at the start and end of each analysis and after every 12 unknowns [20,42,43,64]. The precision on laboratory standards is <0.15‰.”
Remark g. It is not very clear how Fig. 2 was obtained. In particular how the age of each sample submitted to d13C analyses was estimated? If it was done by using the age model described before then the comment about uncertainty should be considered.
Response to remark g: please see our previous response to your remark b
Remark h. The discussion about the correlation between the measured isotopic data and the literature rainfall records seems to me quite weak. In particular considering that only agreement index (R) is reported but to an in depht analysis is missing. For instance by putting the data on the same graph.
Response to remark h:
Rainfall in this region of southern Africa is derived from convective storms that have particularly localised effects. In order to compare our record with an instrumental record it would need to be from a very nearby source, and none exists. We do not expect that the correlation with the Therrell record should be highly significant except in the low frequency range that reflects synoptic rearrangements. The high frequency component is driven by very localised effects and there is no anticipation that this will match between 2 records that are hundreds of km apart.
However, we have altered Figure 2 to show on the same graph as you suggested the calibration of the Chapman archive with modern precipitation levels registered for the nearest station located in Maun. However, this instrumental record for rainfall is very short (as is generally the case with southern Arica) and no statistically relevant correlation could be made.
Reviewer 2 Report
This study presents a high resolution stable and radioisotope study on the Chapman baobab in northeastern Botswana to try to reveal the paleoclimate variations in African continents. In particular, this study obtained a ca. 900-Year isotopic proxy rainfall record from northeastern Botswana by means of radiocarbon dating in the absence of direct tree ring chronologies. Reliable carbon isotope data was combined these radiocarbon dating results to reveal and rainfall and precipitation events using biweight mean statistical method. This manuscript is well written with reliable stable and radioisotope data and advanced analytical techniques, making them deserve publication in the journal of Forests. Before formal publication, there is a small question that want the authors to clarify. In Table 1, sample CH6-1295 has an older age than sample CH6-1420, but the former has a thinner depth than the later. This means the ages are not consistent their depths, please explain why?
Author Response
Point-by-point response to Reviewer 2
Comment of Reviewer 2:
This study presents a high resolution stable and radioisotope study on the Chapman baobab in northeastern Botswana to try to reveal the paleoclimate variations in African continents. In particular, this study obtained a ca. 900-Year isotopic proxy rainfall record from northeastern Botswana by means of radiocarbon dating in the absence of direct tree ring chronologies. Reliable carbon isotope data was combined these radiocarbon dating results to reveal and rainfall and precipitation events using biweight mean statistical method. This manuscript is well written with reliable stable and radioisotope data and advanced analytical techniques, making them deserve publication in the journal of Forests. Before formal publication, there is a small question that want the authors to clarify. In Table 1, sample CH6-1295 has an older age than sample CH6-1420, but the former has a thinner depth than the later. This means the ages are not consistent their depths, please explain why?
Response to Reviewer 2:
We thank the reviewer for evaluating our manuscript and for the appreciation of our study.
The apparent age inversion should be considered in the light of the calibration. The younger of the two samples (1295) has a 1-sigma minimum age of 1223 CE, while the older sample (1420) has a maximum 1-sigma age of 1154 CE. This means that, at the 1-sigma confidence level, the calibrated dates are consistent with 1295 being younger than 1420 by up to 70 years.
Reviewer 3 Report
Dry regions of the tropics present a high climatological risk of desertification and southern Africa is particularly vulnerable. In the absence of long instrumental records paleoclimate studies are required to reveal the mechanisms of intrinsic changes in the climate system. In this paper, a high resolution climate archive was reconstructed based on carbon isotope analysis and radiocarbon dating of the Chapman baobab in northeastern Botswana. This article has significant scientific significance and the method is appropriate. It needs to be modified appropriately before acceptance.
1. The abstract section is more descriptive of the results and it is recommended that the authors add the scientific questions and significance addressed by this study.
2. Line159-161: What are the mean annual and monthly temperatures in this study area? Please add some more information about the temperature.
3. Line 191: What is the distribution of Chapman baobab? What are the differences in the distribution of the same tree species from other regions? Please add relevant information for further community information.
4. Line236-237: How does the age model derived from this study minimize model error range? Approximately how much did the longer time series results keep the model's error under control?
5. Line 370-390: The correlation between CH3 and ENSO is missing in this section. The specific reasons for the phase change of the correlation and the reversal of the situation mentioned in the section are not yet clear, but several possible influencing factors can be suggested based on previous studies.
6. Line 425-430: Comparison with modern precipitation levels can be presented with figure so that readers can understand the specific differences.
Author Response
Point-by-point response to Reviewer 3
General remark:
Dry regions of the tropics present a high climatological risk of desertification and southern Africa is particularly vulnerable. In the absence of long instrumental records paleoclimate studies are required to reveal the mechanisms of intrinsic changes in the climate system. In this paper, a high resolution climate archive was reconstructed based on carbon isotope analysis and radiocarbon dating of the Chapman baobab in northeastern Botswana. This article has significant scientific significance and the method is appropriate. It needs to be modified appropriately before acceptance.
Response to general remark: Thank you for agreeing to evaluate our manuscript, for highlighting its importance and suggesting improvements.
Remark 1. The abstract section is more descriptive of the results and it is recommended that the authors add the scientific questions and significance addressed by this study.
Response to Remark 1: The abstract was altered to include the phrase “The wettest conditions of the last millennium were registered before 1450 while the driest period occurred in 1835.” The significance of the study was clearly stated at the end of the abstract by the following addition “The results contribute to a better understanding of the past climate of southern Africa for which paleoclimate reconstructions remain scarce.”.
Remark 2. Line 159-161: What are the mean annual and monthly temperatures in this study area? Please add some more information about the temperature.
Response to Remark 2: We have introduced in the text more information on temperature as follows: “Satellite data indicates a mean annual temperature of 35.2 °C for the area and a mean temperature during summer months of 39.7 °C.”
Remark 3. Line 191: What is the distribution of Chapman baobab? What are the differences in the distribution of the same tree species from other regions? Please add relevant information for further community information.
It is necessary to explain that Chapman’s Baobab is just a name give to a particularly spectacular example of Adansonia digitata L.
Response to Remark 3: It is absolutely necessary to stress that the paleoclimate reconstruction is based on samples from the Chapman baobab. This particular baobab is a historic tree that was a national icon of Botswana, heavily visited by tourists. In 2019, we published a paper on the age and architecture of the Chapman baobab, also investigating the reasons for its demise (see our reference 63. Patrut, A.; Woodborne, S.; Patrut, R.T.; Hall, G.; Rakosy, L.; Winterbach, C.; von Reden, K.F. Age, growth and death of a national icon: The Historic Chapman Baobab of Botswana. Forests 2019, 10(11), 983. https://doi.org/10.3390/f10110983).
I cite from this published paper: “The baobab was named after the South African explorer and hunter James Chapman, who visited the tree on July 10, 1852 [9,10]. Recent research indicates that the tree had been previously visited in 1849 by David Livingstone, the Scottish missionary and explorer [11].”
The demise of the Chapman baobab was reported earlier in a different paper (see our reference 59. Patrut, A.; Woodborne, S.; Patrut, R.T.; Rakosy, L.; Lowy, D.A.; Hall, G.; von Reden, K.F. The demise of the largest and oldest African baobabs. Nature Plants 2018, 4, 423-426. https://doi.org/10.1038/s41477-018-0170-5), together with the deaths of other monumental baobabs in southern Africa. Thus, as is the case with monumental baobabs, the name of the tree has existed for generations and it is important for researchers to keep the name of the tree.
Large African baobabs are rather an exception than the norm. The Chapman baobab was a solitary individual on a radius of about 10 km. Our previous papers mention that it was relict tree. The African baobab is vastly distributed in Sub-Saharan Africa and ecological niche modelling results indicated that sensitivity to frost, seasonal variation in temperature and rainfall are factors that determine 55.1% of its distribution in East Africa (see Cuni Sanchez A., Osborne P.E., Haq N., Identifying the global potential for baobab tree cultivation using ecological niche modelling, Agroforestry Systems 80(2): 191-2010, 2010, doi: 10.1007/s10457-010-9282-2.).
Given that the African baobab has a vast distribution, that its recruitment next to settlements was assisted by humans for millennia, and that is was successfully introduced in the subtropics worldwide, differences in its distribution are not particularly relevant. Furthermore, baobab distribution was not an objective of our study.
Remark 4. Line236-237: How does the age model derived from this study minimize model error range? Approximately how much did the longer time series results keep the model's error under control?
Response to Remark 4:
The age model does not presume to minimise age errors that are set by both the precision of the radiocarbon dates, and the assumptions made in applying a linear radial growth rate. The approach has the advantage that, assuming there should be one consistent d13C record, the dates from both stems can be used to constrain the d13C record. This helps to reduce the probability options in the calibration of radiocarbon dates.
We are uncertain what is being asked exactly about the effect of the longer core. For both stems the maximum ages are approximately the same. While CH3 was 1.06 m and CH6 was 1.47 m they were dated with 9 and 10 14C analyses respectively. Because the dates for both stems constrain the d13C time series, there is no dominant effect of one core over the other in the formulation of the age model.
Remark 5. Line 370-390: The correlation between CH3 and ENSO is missing in this section. The specific reasons for the phase change of the correlation and the reversal of the situation mentioned in the section are not yet clear, but several possible influencing factors can be suggested based on previous studies.
Response to Remark 5: Thank you for highlighting this error. We did not make correlations for CH3 and CH6 individually, we combined both archives to produce a single isotopic time series. We have rectified the error in the text.
The reversal was also reported by Woodborne et al. 2016 in southern Africa, although about 5 decades later (see subsection 3.1.).
We have added the following phrase in the subsection: “It is possible that the Pacific Decadal Oscillation plays a role in the distribution of dry–wet changes associated with ENSO, but the combination is not yet clearly understood” This is supported by the new reference Wang, Shanshan, Jianping Huang, Yongli He, and Yuping Guan. Combined Effects of the Pacific Decadal Oscillation and El Niño-Southern Oscillation on Global Land Dry–Wet Changes. Scientific Reports 4, no. 1 (2014). https://doi.org/10.1038/srep06651
Remark 6. Line 425-430: Comparison with modern precipitation levels can be presented with figure so that readers can understand the specific differences.
Response to Remark 6: As suggested, we have altered Figure 2 to show the calibration of the Chapman archive with modern precipitation levels registered in the nearest station (Maun). However, the instrumental record for rainfall in Botswana is very short and no statistically relevant correlation can be made.
Round 2
Reviewer 1 Report
All the issues I raised have been addressed. The paper can be accepted.
