the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Evaluating the Tea Bag Index approach for different management practices in agroecosystems using long-term field experiments in Austria and Sweden
Maria Regina Gmach
Martin A. Bolinder
Lorenzo Menichetti
Thomas Kätterer
Heide Spiegel
Olle Åkesson
Jürgen Kurt Friedel
Andreas Surböck
Agnes Schweinzer
Taru Sandén
Abstract. Litter decomposition is an important factor affecting local and global C cycles. It is known that decomposition through soil microbial activity in ecosystems is mainly influenced by soil type and climatic conditions. However, for agroecosystems, there remains a need for a better understanding how management practices influence litter decomposition. This study examined the effect of different management practices on decomposition at 29 sites with long-term (mean duration of 38 years) field experiments (LTEs) using the Tea Bag Index (TBI) protocol with standard litter (Rooibos and Green tea) developed by Keuskamp et al. (2013). The objective was to determine if the TBI decomposition rate (k) and stabilization factor (S) are sensitive enough to detect differences in litter decomposition between management practices, and how they interact with edaphic factors, crop type and local climatic conditions. Tea bags were buried and collected after ~60 and 90 days in 16 Austrian and 13 Swedish sites. The treatments at Austrian LTEs focused on mineral and organic fertilization, tillage systems and crop residues management, whereas the Swedish LTEs addressed cropping systems, mineral fertilization and tillage systems. The results showed that in Austria, decomposition differed more between sites than between treatments for the same experiment category. Incorporation of crop residues and high N fertilization increased k. Minimum tillage had significantly higher k compared to reduced and conventional tillage. In Sweden, litter decomposition differed more between treatments than between sites. Fertilized plots showed higher S than non-fertilized and high N fertilization had the highest k. Growing spring cereal lead to higher k than forage. Random Forest regressions showed that k and S were mainly governed by climatic conditions, which explained more than 70 % of their variation. However, under similar climatic conditions, management practices strongly influenced decomposition dynamics. Thus, the TBI approach may be suitable to apply in a more large-scale network on LTEs for evaluating decomposition dynamics more precisely.
- Preprint
(1654 KB) - Metadata XML
-
Supplement
(280 KB) - BibTeX
- EndNote
Maria Regina Gmach et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-1229', Moritz Laub, 01 Sep 2023
I want to congratulate the authors on an impressive study that took a lot of samples and generated a very rich dataset. I think that the questions posed in the introduction are interesting and of scientific relevance. Due to the large dataset and the many interesting conclusions that could be drawn from that, I do think it should eventually be published in SOIL. However, I think that the study needs to be improved in a number of aspects, before it can be accepted for publication. Also, I think the study focuses too much things that are well known (effect of temperature and moisture) and only superficially touches the more interesting aspects (effect of other soil properties such as SOC, CN – and that of management). The authors could probably make more of this highly interesting dataset, if the methods were refined. I specified a few suggestions on how that could be done, below.
Comments to specific parts:
Introduction:
Overall well written – not much to improve there.
Just one thing: The term fertilization should be replaced by fertilizer application (throughout the text), because fertilization refers to a specific mechanism in sexual reproduction and is just not the right word.
Methods:
The methodology lacks some important details, especially on the rate modifiers from ICBM, on the way statistics were applied and on dataset split for different forms of analyses. Also, with the high number of experiments, it is especially important to structure their description and the results by some common denominator across experiments. The way it is now, I find it very hard to follow.
I am further wondering, if you could not make much more of the data if you included a temperature function (and maybe moisture function) in the TBI calculation directly (e.g. W(t)= a e-k t f(T) + (1-a), where f(T) is a temperature function as used in most biogeochemical models. That may help you to distill out better the effect of management. It is well known that that temperature and moisture is important, so I would think that the question what is important for a temperature and moisture adjusted k and S of the TBI would be of much higher scientific relevance. After all models such as ICBM are around for many decades and have such rate modifiers since long.
I found it a bit hard to follow all the different experiments and how they are structured. The tables help to understand them but the text is a bit unstructured. It seems a bit random what details are described about each of them vs what details omitted. I would suggest you either make the description less details and rely fully on the tables or you describe every experiment in the exact same order (e.g., 1) cropping system, 2) tillage system 3) inputs 4) exports etc.).´
I think that statistical tests using ANOVA are not correctly accounting for the fact that there are a lot of samples in the dataset that are not independent (e.g. from different years at the same sites). I would think that a mixed linear model would be much more suitable, given such nested structure of data.
I am a bit concerned that samples from Sweden and Austria were treated differently (measuring ash content versus not). This means that the results are not comparable – at minimum this difference in treatment should be a covariate in random forest. Also, for Austria data – how can you assure that differences, especially those between sites, but due to soil structure maybe even those between treatments are not just more or less soil particle sticking to the tea?
To me, it is not really clear how Retemp and Rewat were calculated and used in this study. This is poorly described. Did you actually run ICBM or did you just use the typical Retemp and Rewat functions and calculate them yourself? As specified above, I think this is actually a good idea, you just need to make it more clear what it is that you actually did.
Results:
The results are also presented in a somewhat difficult to follow fashion. With so many abbreviations, it would help to write out at least the treatments instead of describing K, FW, CRR… And then you talk about AT1, AT2 ect. and suddenly about the Fuchsenbiggl LTE – this is also confusing. Also, is it not clear if all the differences you describe are significant – and Fig 2 and 3 also do not show this.
Discussion:
In the discussion, I am not sure if the comparison of differences in k between different sites is all that meaningful. As long as k and S are not moisture/temperature corrected those have probably the highest influence and the rest is mostly speculation. Also I would start with the major findings of the RF model (what are the most important factors determining k and S) before going into the details of each individual experiment.
Line specific comments:
L44 Were the random forest regressions done by country or with the joint dataset?
L58 I would rather call it emission capacity, because as far as I know, most agricultural soils are actually loosing SOC.
L62-64 Maybe refine this statement a bit. Microbes are actually a main contributor to stable SOC (https://www.nature.com/articles/s41561-022-01100-3), wheres mere C input is not stable at all.
L70 What is exactly is meant by “extended process”? And why is it only decomposition not stabilization? (The focus on decomposition is very teabag focused)
L85 I think it would be suitable to display the differential equation used to derive S and k and to shortly describe the fitting process
L95 I would argue that, because TBI is not normalized to mean annual temperature, it is more influenced by temperature than soil quality.
L96 The argument that agroecosystems have the least been studied does not match the fact that you have much more references there???
L102 I think it would be good to elaborate a bit more what LTEs were used. E.g., similar soil conditions, different management.
L107 distinguish instead of detect?
L120/121 Please clarify: What are C balance practices? Residue removal vs incorporation? And how can an LTE be used to study soil fertility if treatments have the same soil?
L163 Please clarify if the dairy production treatments also include cash crops
L171 You have not described the rotation in detail, so the specifics for some crops are a bit confusing here.
L205 What do you mean by identified? Uniquely labeled? And was there only one bag buried in each treatment at each site?
L212 Did you assure in any other way that no soil stuck to the tea litter if ash content was not determined?
L214 The four bags for AT sites were not described above. Or did I miss it?
L218 So for SE sites you did determine soil contamination? I find it a bit problematic for comparability that this was done for SE but not AT.
L227ff It is not clear which measurement was done at which level of detail. I would suspect TOC at the plot level, texture at experiment level, maybe – but please specify.
L238 You have repeated measurements in your data. How did you assure that by this you did not artificially inflate statistical power? Have you averaged repeated measurements? Also did you average the 4 replicates per treatments? I think you need to do this, or you use a mixed linear model, probably best with a nested structure such as experiment/block/treatment
L242 do you mean you adjusted your k by it? I am confused because the whole article is about litter and now you talk about SOC pools.
L243 It is probably just a rate modifier, i.e., f(T) * f(m), so actually just a function of temperature and moisture. Arguably those two are influenced by climate, crop and soil properties. Maybe best just give the function of the rate modifiers
L247 Should Retemp not be a function of soil temperature?
L253 Which variables? What do you mean by “more accurate results”?
L256ff I think you can shorten the RF description to the most essential and just cite appropriate literature
L266 I think you should cite some paper describing how “node purity” and “Gini index” are defined
L275/6 What is the difference between TN and N? MATTBI should be MTTBI no?
L303 So is AT5 and AT6 actually the same experiment just in different years?
L339 You have not described this decomposition model in the methods.
L355 Fewer than what?
L359-61 Yes, and this is exactly why I would advocate that you also try the RF on a temperature and moisture corrected k and S value from TBI, e.g. by using Retemp and Rewat. Because the finding that temperature and moisture are important is really nothing new, but the influence of management factors and soil types much more interesting.
L430 – how much ash was actually there in Sweden? Does is give you an indication that your AT data may be strongly biased? How much is the ash content of the litter before burial?
L438 I would start the discussion with this section
L445 I would argue that precipitation and Rewat are likely strongly correlated. So if you have both in your dataset, you cannot conclude that precipitation is not important if not selected by RF.
L458/9 I do not understand what you want to express here.
L462 Where in your results do you display node purity? Have I missed it?
L485 To me, this is the most interesting part of your results. You should put more emphasis on this and less on the well-known temperature/moisture effects. It would also be interesting in which direction SOC, CN etc influences S?
Table 4: I think you should state mean k and S per country and tea type and also give measures of uncertainty (e.g. SD).
Fig 2: The font size is very small and hard to read.
Fig 2 and 3 should further indicate which treatments differ significantly between treatments at the same site (this is only in the text, so far).
Citation: https://doi.org/10.5194/egusphere-2023-1229-RC1 -
RC2: 'Comment on egusphere-2023-1229', Tatiana Elumeeva, 08 Sep 2023
The study “Evaluating the Tea Bag Index approach for different management practices in agroecosystems using long-term field experiments in Austria and Sweden» investigates the effects of various types of management on decomposition processes estimated by the recently widely used TBI approach. This is an attempt to generalize a lot of field experiment data from two regions and with different kinds of experiments (organic and mineral fertilization, cropping and tillage systems) of different duration and under different crops. The structure of initial data complicates the analysis and further presentation of the results, and sometimes the text is difficult to follow, especially when the authors refer to site abbreviations AT1, AT2, etc. I think the use of multivariate methods would be useful for visualization of the results, may be the PCA ordination of sites (treatments) using soil and climate variables to reveal the main gradients of environment, and highlighted types of treatment and directions of STBI and kTBI changes.
Below are the minor comments on the text:
Line 95: “Camellia sinensin” – Camellia sinensis
Line 97: “Aspalanthus linearis” replace by Aspalathus linearis
Lines 238–240: It would be better to describe the ANOVA a bit more. Here it seems the ANOVA was applied to the total data set, but when reading Table S2 it seems that Tukey’s test was applied mostly to the single sites and to the 2 or 3 sites with the same treatments in Austria and groups of experiments in Sweden.
Line 303: “Comparing years for the same experiment type, AT5 (2015) had higher S than AT6 (2016).” – is that due to years or due to spatial heterogeneity between sites?
Lines 482-483: “The stabilization factor S expresses the degree by which the labile fraction of the plant material is decomposed.” – Doesn’t it expresses the degree by which the labile fraction was transferred to recalcitrant one?
Line 802: “IF-K with NP: K inorganic fertilization with 120 kg N ha-1;” – What about P?
Figures 2 and 3 contain too many values, so they are difficult to understand and unclear, which values should be compared.
The following citations are absent in the list of references:
Line 97: Pino et al. 2021
Line 125: Spiegel et al., 2018; Lehtinen et al., 2017
Line 126: Lehtinen et al., 2014
Line 127: Spiegel et al., 2007
The following references are not cited in the text:
Line 537: Al-Kaisi et al., 2008.
Line 593 Cornwell et al, 2008.
Line 600: Couteaux et al, 1995.
Line 608: Didion et al., 2016.
Line 614: Domínguez et al., 2014.
Line 620: Dubeux Jr et al., 2006.
Line 623: Duddigan et al. 2020.
Line 626: Elumeeva et al., 2018.
Line 636: Food and Agriculture Organization (FAO), 2005.
Line 643: Freschet et al., 2012.
Line 654: IPCC 2021
Line 654: IPCC 2022
Line 687: Kätterer et al., 2014.
Line 697: Kohmann et al., 2019.
Line 714: Martin et al., 2020.
Line 733: Pingel et al., 2019.
Citation: https://doi.org/10.5194/egusphere-2023-1229-RC2 -
CC1: 'Comment on egusphere-2023-1229', Taiki Mori, 13 Sep 2023
As a premise for evaluating this paper, it is important to note that, in the context of the TBI approach, the parameter k is defined as the decomposition constant characterizing the asymptote model describing the decomposition curve of rooibos tea. Meanwhile S is computed as the ratio of the stabilized fraction to the hydrolysable fraction in green tea. TBI approach offers the advantage of determining both k and S using a single set of mass loss values acquired during an incubation period of approximately 90 days.
Importantly, this approach assumes two essential assumptions, and only when both of these assumptions are met, the need for time-series data is obviated. First, the most portion of the hydrolysable fraction in green tea is decomposed within the initial 90 days (the first assumption). Secondly, the stabilization factor S, denoting the ratio of the stabilized to the total hydrolysable fractions, remains the same for rooibos tea as it does for green tea (the second assumption).
Given the aforementioned premises, my first comment on the authors' work pertains to their utilization of time-series data within their research. Despite the acquisition of these time-series data in their work, the authors omitted a rigorous assessment of their alignment with the fundamental assumptions underpinning the TBI approach. Instead, the authors simply computed the parameters k and S following the TBI method. I strongly recommend a thorough examination to ensure that these data comply with the specified assumptions.
Moving on to my second point, it is imperative to note that these two premises have been challenged by previous works. My own research demonstrated that the application of an incubation study revealed a failure to meet these assumptions. This outcome holds particular significance (more robust than field study) due to the control of temperature conditions maintained throughout the incubation period (Mori 2022a). One could posit the argument that the TBI approach retains its utility for the comparative analysis of the "relative decomposition rate" across diverse ecosystem (or soil) types. However, another paper demonstrated the absence of a positive correlation between TBI-derived parameter k and the k values established through time-series data (Mori 2022a). Consequently, within the context of scientific research, it becomes apparent that the TBI approach may not be considered a suitable methodology.
Mori, T. (2022a). Validation of the Tea Bag Index as a standard approach for assessing organic matter decomposition: A laboratory incubation experiment. Ecol. Indic., 141, 109077.
Mori, T. (2022b). Is the Tea Bag Index (TBI) Useful for Comparing Decomposition Rates among Soils? Ecologies, 3, 521–529.
I acknowledge that a substantial number of publications employ the TBI protocol. Therefore, I refrain from endorsing the outright rejection of the publication of this paper. However, I do propose, as a minimum, two recommendations: (i) refraining from combining data derived from 60 days and 90 days when calculating TBI (mentioned in L271), and (ii) clearly mentioning the aforementioned limitation of the TBI method (i.e., the accuracy of the TBI method is challenged) in the paper to prevent any potential misinterpretation or misunderstanding by readers.
I would also propose refraining from constructing a graphical representation illustrating the relationship between parameters k and S (FIGURE 4). This recommendation is grounded in the outcomes of a simulation study conducted by Mori et al. (2022), which demonstrated a kind of autocorrelation between k and S.
Mori, T., Nakamura, R. & Aoyagi, R. (2022). Risk of misinterpreting the Tea Bag Index: Field observations and a random simulation. Ecol. Res., 37, 381–389.
Citation: https://doi.org/10.5194/egusphere-2023-1229-CC1 -
RC3: 'Comment on egusphere-2023-1229', Ute Hamer, 17 Sep 2023
The authors present an interesting, valuable, large dataset on decomposition of rooibos and green tea in different long-term agricultural field experiments across Austria and Sweden. It is an important topic to find indicators suitable to assess the influence of different agricultural management practices on soil quality. The manuscript is well written and, in most points, easy to follow. Considering some aspects in more detail, could help clarify main conclusions.
Especially in the Swedish dataset, there seems to be more information available regarding this topic, because decomposition was measured after 15, 30, 60 and 90 days. Thus, k values can be modelled individually per site via a decay function as was done in Figure S1 for all sites together. But I strongly recommend doing this analysis separately per site. See papers suggested already by Mori and also our study Middelanis et al. 2023.
To combine this with suggestions of reviewer 1 would be ideal to focus more on management effects instead of climate effects. Instead of showing general differences between Austria and Sweden in Figure 4 it would be more interesting to focus on management aspects such as fertilizer application, tillage system ect.
In general, I find it hard to understand why you mixed up data obtained after 60 days of decomposition and those obtained after 90 days. The labile fraction of rooibos tea is not yet decomposed (even after 90 days), as you show in Figure S1. To my understanding this does not make sense. But you spend much space in your manuscript to present these data. What is your conclusion based on the combined data analysis versus the individual data analysis?
I miss a bit the discussion on the suitability of the random forest analysis and the conclusions which can be drawn from the results. Does it really make sense to include 17 variables as in Figure 5 b, 5 d and 5e? As far as I see in the Figure R² increases only marginally. What is the lesson which can be learned from the information that the factor treatment was selected as e.g. the 7th variable out of 17 explaining S after 60 days (Figure 6d)? When I looked up the factor treatment in the Material and Method section it notes that there are 30 levels of treatment (line 281), which ones? For your first analysis you distinguished between CMP, ROT and TS. Why do you not consider this in your random forest models?
Line 281: Why is maize not included in the crop factor of your random forest model?
I find it quite interesting that “In Sweden, litter decomposition differed more between treatments than between sites” (line 41, 42), although the gradient in soil properties seems to be larger in Sweden than in Austria. According to Table S1 clay content in Sweden ranges between below 5 to 50 %. Would be quite interesting to figure out under which circumstances treatment effects become more visible. Do you have an explanation for this?
Table S1: Please add information on soil type, as indicated in the heading! To which soil depth do the data refer to? And to which treatment? Sites have different treatments and addition of FYM might for example increase SOC… Please, add more precise information for all treatments of specific sites. Please also check soil texture and clay content, e.g. a silty clay with 5.6 % clay at SE 4 seems to be strange.
Line 107 to 109: check sentence structure.
Line 204f: how many tea bags have been burrowed per site? Only 1 green and 1 rooibos tea bag? Please indicate.
Line 234: was the pH in Swedish soil samples also measured in 0.01 M CaCl2?
Line 303f: how can you be sure that this is a year effect and not an effect of the crop species?
Line 402ff: again: how can you be sure that differences are due to climatic conditions and not to different effects of maize versus wheat (there is a lot of literature around showing beneficial effects of wheat compared to maize)
Line 456ff: this is rather a repetition of results than a real discussion.
Table 1: It is a bit confusing using different site numbers for the same site in different years with a different crop in rotation! I suggest to number according to locations and year, eg. AT1_1 and AT1_2 for location AT 1 and year 1 and year 2, respectively.
Figure 3: why are Whiskers missing in some cases, is n too small? Please indicate the number of replicates per treatment!
Middelanis, T., Pohl, C. M., Looschelders, D., & Hamer, U. (2023). New directions for the Tea Bag Index: Alternative teabags and concepts can advance citizen science. Ecological Research, 1–10. https://doi.org/10.1111/1440-1703.12409
Citation: https://doi.org/10.5194/egusphere-2023-1229-RC3
Maria Regina Gmach et al.
Maria Regina Gmach et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 284 | 56 | 14 | 354 | 27 | 3 | 3 |
- HTML: 284
- PDF: 56
- XML: 14
- Total: 354
- Supplement: 27
- BibTeX: 3
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1