| 23 Dec 2025
Estimation of the degree of decomposition of peat and past net primary production from mid-infrared spectra
Abstract. The degree of decomposition of peat (γ) is useful to understand peatland degradation and peat accumulation, to reconstruct past net primary production (NPP), and to improve peatland models. None of the available decomposition indicators allows to estimate γ with sufficient accuracy. We suggest prediction of γ measured in litterbag experiments from mid-infrared spectra (MIRS) as a novel decomposition indicator, γMIRS, and compute prediction models for γMIRS with available litterbag experiments and litter data from diverse species from the Peatland Mid-Infrared Database. For individual litter samples, the prediction models fit the data well, have reasonable prediction errors (average RMSE between 0.09 and 0.12 g g−1), and neither confound differences in litter chemistry nor differences in silicate contents with decomposition losses. We show that an underestimation of γ by γMIRS matches theoretical expectations; it can therefore be compensated, using plant macrofossil analysis data as a first approximation to mass fractions of peat components and a simple mixing model, or it can be avoided with component-specific measurements instead of bulk measurements. This allows to estimate γ of peat samples and of dominant litter types and therefore also to reconstruct past NPP. To illustrate the approach, we analyze three cores from European mountain bogs and discuss how it can be used to improve process models and support restoration of peatlands. In particular, we test previously suggested relations between the saturated hydraulic conductivity and γ, illustrate how γ measured on individual litter types may allow to use peat cores as natural litterbag experiments, and define reference states for γ and NPP for the three analyzed peat cores. Improvements to reduce prediction errors of the approach require more diverse litterbag data, especially woody species and more decomposed litter. Further improvements can be achieved with measurements of MIRS on individual macrofossil types instead of bulk measurements, and an improved estimation of mass fractions of macrofossil types in peat samples instead of assuming that macrofossil abundances equal macrofossil masses.
The manuscript by Teickner et al. describes a method using mid-infrared (MIR) spectroscopy to estimate the degree of decomposition (defined as the fraction of initial mass lost due to decomposition) and to derive net primary production (NPP) from this metric. The approach is based on MIR analysis of vegetation and litter samples from litter-bag decomposition experiments, which are used as training data to develop models subsequently applied to peat cores. The aim is to assess whether reconstructed degrees of decomposition and NPP are consistent with independent proxies derived from organic-matter composition, vegetation, and degradation indicators. In addition, the authors apply the model to previously proposed relationships between saturated hydraulic conductivity and degree of decomposition. Three different models are developed, and several tests are performed to evaluate, among other factors, the influence of varying litter proportions and the presence of siliceous minerals in peat samples. The manuscript is well written, the figures are of high quality, and the overall structure of the text is clear and appropriate.
One of the main limitations of the approach, which is also acknowledged by the authors, is that the training dataset is dominated by vegetation samples (i.e. samples with a degree of decomposition equal to zero), while only a limited number of samples represent degraded material derived from litter-bag decomposition experiments. As the authors correctly note, peat degradation is a complex process influenced by multiple factors, including vegetation type and plant components, oxygen availability, and changes in water-table depth. Consequently, peat—and especially peat cores—represents a highly complex matrix, making it inherently challenging to quantify how much organic matter has been lost relative to the original composition, which itself is unknown, using an MIR model trained primarily on vegetation samples.
Despite this limitation, the authors make a substantial effort to demonstrate the potential of the approach, and the comparisons between model outputs and palaeo-reconstructions from peat cores yield promising results. From this perspective, the manuscript represents a valuable contribution as a first step toward estimating degree of decomposition and NPP using relatively simple measurements and models such as MIR spectroscopy. The study will likely be useful for researchers interested in developing and refining similar approaches and in advancing this line of methodological research.
However, given the current limitations, these should be more clearly highlighted in both the title and the abstract, for example by framing the study explicitly as a “first step” or “proof-of-concept.” In addition, it should be emphasized in the abstract, discussion, and conclusions that this method alone does not yet appear sufficiently robust to be applied independently, and that it should be used in combination with more established palaeo-reconstruction approaches (e.g. macrofossil analysis). The use of this approach for other research will demonstrate its usefulness and accuracy in the future.
I strongly suggest including a schematic or conceptual figure that clearly explains the overall modeling workflow, from the training samples to the final results, including the different tests performed. Such a figure should indicate at which stages MIR analyses are conducted, as well as the input and output data at each step. This would greatly improve the clarity of the methodological approach for the reader.
The description of the methods could also be improved. For example, the main text should provide more detailed information on the litter-bag experiments used in the training dataset, including the number of samples, incubation durations, and the specific time points at which samples were analyzed by MIR. In addition, the manuscript frequently refers to previous studies by the authors, with which readers may not be familiar. I recommend providing more explicit descriptions rather than assuming prior knowledge—for instance, clearly explaining what the pmird database is and whether the MIR data from the peat cores are included within this framework.
Did the authors take any measures to check whether there were any differences between the instruments used to measure the MIR? Quality control?
With respect to nitrogen content, it would be helpful to explicitly state how many samples have measured nitrogen concentrations and how many values are predicted. The authors could also discuss the implications of this “double modeling” approach, in which nitrogen is first predicted and then used for comparison with degree of decomposition estimates that are themselves derived from MIR data.
Peat samples represent a complex mixture of organic matter derived from vegetation at different stages of decomposition, as well as microbial biomass. This complexity can influence the molecular signals measured in peat, as classically demonstrated by Chapman et al. (2001). This point may be relevant to address more explicitly in the discussion, as some MIR bands measured in peat samples may partly reflect microbial-derived signals rather than plant-derived organic matter alone.
It is somewhat surprising that no evaluation or comparison between the obtained degradation index and the C/N ratio is presented, despite the C/N ratio being widely used as an indicator of mass loss in peatlands and being relatively easy to measure. I suggest that the authors include such a comparison, as it could be particularly useful when evaluating model results and for further studies, at least for those peat cores where C/N ratios are available.
In the discussion, I also miss a more quantitative treatment of models “over- or underestimation”. Including percentages or value ranges would strengthen the discussion and allow a clearer assessment of model performance.
There are some issues with the references that should be addressed. Two references appear with the same authors (“Teickner, H. and Knorr, K.-H.”) but different titles, and a similar issue occurs for “Teickner, H., Pebesma, E., and Knorr, K.-H.” Please ensure that each in-text citation can be unambiguously linked to a single reference in the bibliography. One of these references is a preprint, which should be considered carefully in the context of the publication process.
Finally, I recommend rephrasing the end of Section 3.8. The text is somewhat wordy, and the long parenthetical sentences disrupt the reading flow.