Quantifying meteorological impacts on local landfill methane emission by using field measurements and machine learning

Kim, Donghee; Jeong, Sujong; Chang, Dong Yeong; Joo, Jaewon

doi:10.5194/egusphere-2025-3369

Preprints

https://doi.org/10.5194/egusphere-2025-3369

Preprints

15 Sep 2025

| 15 Sep 2025

Quantifying meteorological impacts on local landfill methane emission by using field measurements and machine learning

Donghee Kim, Sujong Jeong, Dong Yeong Chang, and Jaewon Joo

Abstract. Landfills are a major anthropogenic source of methane (CH₄), contributing up to 20 % of global CH₄ emissions. Although CH₄ emissions from landfills are highly sensitive to meteorological conditions, their response to climate variations remains poorly understood, leading to substantial uncertainty in emission projections under climate change. This study evaluated the impact of meteorological factors on landfill CH₄ generation, using a site-specific machine-learning-based model optimized for temperature and precipitation. The model optimized for meteorological conditions performed better than conventional models such as LandGEM and the IPCC model, with a root mean squared error (RMSE) of 6.57 million m³CH₄, a mean absolute error (MAE) of 4.91 million m³ CH₄, and Pearson correlation coefficients of 0.89, when compared with field measurements. CH₄ generation exhibited a linear correlation with increasing temperature, and a parabolic response to increasing precipitation. Quantification of the contributions of the meteorological variables, revealed that temperature accounted for 5.96±3.06 %, and precipitation for 7.38±0.58 % of the total modeled CH₄ generation. These results highlight the high importance of incorporating meteorological variability into landfill CH₄estimation to improve predictive accuracy, and emphasize the need of stronger and faster CH₄ mitigation efforts under climate change.

Received: 13 Jul 2025 – Discussion started: 15 Sep 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Donghee Kim, Sujong Jeong, Dong Yeong Chang, and Jaewon Joo

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-3369', Anonymous Referee #1, 03 Nov 2025
This is a well-prepared and timely study addressing an important topic in methane emission quantification. The authors present a clear, comprehensive, and technically rigorous analysis of meteorological impacts on landfill methane generation using a machine-learning–based approach (CLEENopt) optimized with field measurements. The paper is logically structured, the methodology is well justified, and the discussion is supported by both empirical data and relevant literature.
Minor suggestions for further improvement:
The Introduction would benefit from a clearer justification of why including meteorological factors (temperature and precipitation) in landfill CH4 modeling is necessary. Although these variables are often cited as key environmental controls on microbial methane production, it remains unclear whether landfill CH4 generation is indeed strongly sensitive to them. Existing models such as IPCC FOD and LandGEM do not explicitly account for meteorological parameters, yet they sometimes perform equally well or even better. Without evidence that temperature and precipitation substantially improve predictive accuracy, the rationale for incorporating them is not fully convincing at this stage.

Consider a more explicit discussion on potential transferability of CLEENopt to other climatic or waste management contexts. While the model is convincingly optimized for the SLS, it would strengthen the paper to discuss its applicability to other regions with different climatic regimes, waste compositions, or operational practices. A short evaluation of how the calibration parameters (e.g., temperature, precipitation sensitivity, or waste composition factors) could be generalized, or what site-specific adjustments would be required, could broaden the scientific impact of the study and highlight its potential for international or large-scale applications.

In the Discussion, a short paragraph linking these findings to national inventory improvement or IPCC Tier 2/3 applications could strengthen the applied relevance. Since one of the key motivations of this work is to enhance methane emission estimation accuracy, it would be valuable to explicitly connect the results to national GHG inventory frameworks. For example, discussing how the CLEENopt model could inform refinement of Tier 2/3 parameters under the IPCC guidelines, or contribute to improving uncertainty estimates in landfill CH4 inventories, would clearly position this research within broader policy and reporting contexts.
Citation: https://doi.org/10.5194/egusphere-2025-3369-RC1
RC2: 'Comment on egusphere-2025-3369', Anonymous Referee #2, 03 Nov 2025

Kim et al. present a novel, site-specific machine-learning-based model, termed CLEEN_opt, to quantify methane (CH₄) emissions from the Sudokwon Landfill Site (SLS) in Korea. The authors correctly identify that conventional models, such as LandGEM and the IPCC Waste Model, are insufficient for projecting emissions under climate change, as they largely neglect the impact of meteorological conditions. The manuscript also highlights the limitations of the existing CLEEN model, which, while incorporating temperature and precipitation, requires further site-specific calibration to be applicable to new regions.
The core novelty of this work lies in the development of a calibration methodology that utilizes a Random Forest (RF) machine learning model to generate a site-specific scaling factor, F_RF. This factor is designed to adjust the laboratory-derived first-order decay constant (k_lab) from the original CLEEN model to match field-observed conditions, which are represented by an inversely modeled k_actual. The result is an adjusted, site-specific decay constant, k_adj, which is then used to estimate CH₄ generation.
The manuscript addresses a critical and timely topic, as quantifying non-CO₂ greenhouse gas emissions under climate change is of paramount importance. The methodology for site-specific calibration using field data and machine learning is a valuable contribution to the field.
However, the work in its current form suffers from significant methodological ambiguities and a critical oversimplification of the environmental drivers governing CH₄ generation. The conclusions are therefore not fully supported by the analysis as presented. The manuscript has the potential to be a strong contribution, but only after these core issues are addressed.
Therefore, the recommendation is that a minor revision is required before this manuscript can be further considered for publication. The following are some comments that require attention.
Omission of Synergistic Effects in Meteorological Analysis:

The analysis of meteorological impacts, presented in Section 3.4 and Figure 5, is based on an incomplete and potentially misleading 1-dimensional (1D) sensitivity analysis. The authors examine the effect of temperature (Figure 5a) and precipitation (Figure 5b) independently. The methodology in Section 2.5 confirms this approach, describing scenarios such as "(b) using a fixed mean temperature... and observed precipitation" and "(c) using observed temperature and a fixed mean precipitation".

This "one-at-a-time" (OAT) analytical approach fundamentally prevents the discovery or analysis of interaction effects between the variables. Methanogenesis is a biogeochemical process, and its drivers are not merely additive. The "linear correlation" with temperature shown in Figure 5a is an artifact of averaging across all precipitation conditions; this apparent linearity would almost certainly fail under extreme-dry (desiccation) or extreme-wet (pore saturation) scenarios. A "hot-wet" scenario (characterized by high temperatures and high precipitation) will exhibit a vastly different biogeochemical response than a "hot-dry" scenario (characterized by high temperatures and low precipitation), yet the current analysis cannot distinguish between them.

This flaw raises questions about the subsequent quantification of relative contributions in Figure 6. The analysis improperly partitions the variance from these coupled variables and, by design, ignores the contribution of the interaction term (Temperature x Precipitation), which may be a significant driver in itself.
The authors might need to replace the 1D plots in Figure 5 with a 2-dimensional (2D) sensitivity analysis. A 2D heatmap (e.g., Temperature on the X-axis, Precipitation on the Y-axis, and CH₄generation as the color scale) would be appropriate. This would visualize the true response surface of the model, allowing for a much more robust discussion of the coupled meteorological impacts and identifying the actual optimal and pessimistic conditions that are currently obscured by the OAT analysis.
Minor comments:

Page 1 (Title): "emission" -> "emissions". The plural form is more appropriate as the paper discusses emissions from multiple sources, processes, and sites.
Page 1 (Abstract, L18): "...6.57 million m³ CH₄ a mean absolute error..." -> "...6.57 million m³ CH₄, a mean absolute error..." (A comma is missing in the list of metrics).
Page 1 (Abstract, L22): "...emphasize the need of stronger and faster..." -> "...emphasize the need for stronger and faster..." (Incorrect preposition).
Page 2 (Intro, L7): "...approximately 30% of to global warming..." -> "...approximately 30% to global warming..." (Duplicate word)
Page 2 (Intro, L9): "(IPOC Change, 2007; Prather..." -> This appears to be a significant typographical error. It should almost certainly be "IPCC, 2007; Prather...".
Page 3 (Intro, L2): "...contributing CH₄ emissions..." -> "...contributing to CH₄ emissions...".
Page 4 (Intro, L7): "...LandGEM provides an estimation of..." -> "...LandGEM provides an estimate of..." ('Estimation' is the process; 'estimate' is the resulting value).
Page 4 (Intro, L21): "...with greater accuracy than those of the LandGEM..." -> "...with greater accuracy than that of the LandGEM..." (The antecedent is the singular "accuracy").
Page 6 (Data, L12): "plastic (26.1± 4.7%)" -> "plastic (26.1 ± 4.7%)". Please check for consistent spacing around '±' symbols throughout the manuscript.
Page 7 (Data, L2): "...47.5 m³CH₄I Mg..." -> "... 47.5 m³CH₄ Mg...". (There is a stray 'I' character).
Page 7 (Data, L13): "...simplicity, and flexibility, compared to other..." -> "...simplicity, and flexibility compared to other..." (The comma after "flexibility" is unnecessary).
Page 7 (Data, L17): "...it is estimated based on stable..." -> "...it was estimated based on stable..." (Past tense should be used for actions taken during the study).
Page 8 (Method, L5): The list of six citations for a general statement (Amini et al., 2012; Amini et al., 2013; Lay et al., 1996; Machado et al., 2009; Tolaymat et al., 2010) may be excessive and could be streamlined per journal style.
Page 9 (Method, L15): "...field measurement data has been used..." -> "...field measurement data have been used..." ('Data' is a plural noun).
Page 10 (Method, L10): "...suitable for site-scale monitoring." -> "...suitable for site-scale monitoring." (Hyphenate the compound adjective).
Page 10 (Method, L24): "...its insensitive to outliers." -> "...it is insensitive to outliers." (Missing verb 'is').
Page 10 (Method, L26): "...that is it does not estimate..." -> "...that is, it does not estimate..." (A comma is needed to set off the appositive phrase).
Page 11 (Method, L6): "...waste disposed that entered..." -> "...waste disposed of that entered..." or "...waste that entered...".
Page 11 (Method, L12): "...k_Ial was calculated..." -> "...k_lab was calculated..." (Typo 'Ial' instead of 'lab').
Page 14 (Results, L4): "...value of k_lab calculated using..." -> "...value of k_lab, calculated using..." (The appositive phrase requires a setting comma).
Page 14 (Table 2): The extreme values for k_lab and their corresponding errors (+2585% and +7269%) are a major finding and should be explicitly discussed in the main text of Section 3.1, not just presented in the table.
Page 14 (Results, L22): "...million CH₄m³.r=0.64..." -> "...million CH₄m³, r=0.64..." (An incorrect period(.) is used mid-sentence. It should be changed to comma(,)).
Page 16 (Fig. 4 Caption): "...CLEENopt, CLEEN and actual..." -> "...CLEENopt, CLEEN, and actual..." (Use of a serial comma is recommended for clarity).
Page 18 (Fig. 5 Caption): "...range across all simulated years, and colored shading is the seasonal..." -> This is a run-on sentence. It should be split: "...range across all simulated years. Colored shading represents the seasonal...".
Page 18 (Results, L11): "...thereby gas diffusion..." -> "...thereby inhibiting gas diffusion..." (A verb is missing).
Page 20 (Discussion, L26): "...positive correlation between temperature and CH₄ generations was..." -> "...temperature and CH₄ generation was..." ('Generation' should be singular).
Page 21 (Discussion, L23): "future studies should consider more accurate oxidation rates..." -> This is a key point. The use of a 10% default value is a major assumption and source of uncertainty that warrants more emphasis in the discussion.
Page 22 (Conclusion, L6): "...linear correlation with temperature and a parabolic correlation with precipitation." -> "...linear response to temperature and a parabolic response to precipitation." ('Response' is more accurate than 'correlation' in this context, as it describes a modeled functional relationship, not a statistical correlation).
Page 24 (References): "IPOC Change, 2007" -> This citation is repeated from Page 2. It MUST be corrected to "IPCC, 2007".
Page 27 (References): "Sil, A., Kumar, S., and Wong, J. W.:...model suiting Indian condition..." -> "...model suiting Indian conditions...".
Page 28 (References): "Wang, Y., Pelkonen, M., and Kaila, J.:...Open Waste Manag. J., 5, 2012." -> Page numbers appear to be missing from this journal citation. Please verify.

Citation: https://doi.org/10.5194/egusphere-2025-3369-RC2

Donghee Kim, Sujong Jeong, Dong Yeong Chang, and Jaewon Joo

Viewed

Total article views: 1,855 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
1,757	81	17	1,855	15	19

HTML: 1,757
PDF: 81
XML: 17
Total: 1,855
BibTeX: 15
EndNote: 19

Views and downloads (calculated since 15 Sep 2025)

Month	HTML	PDF	XML	Total
Sep 2025	1,629	7	3	1,639
Oct 2025	66	51	5	122
Nov 2025	62	23	9	94

Cumulative views and downloads (calculated since 15 Sep 2025)

Month	HTML	PDF	XML	Total
Sep 2025	1,629	7	3	1,639
Oct 2025	66	51	5	122
Nov 2025	62	23	9	94

Viewed (geographical distribution)

Total article views: 1,707 (including HTML, PDF, and XML) Thereof 1,707 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 20 Nov 2025

Short summary

This study uses data and machine learning to better estimate methane emissions from a major landfill in South Korea. By considering local weather conditions like temperature and rain, the research improves how landfill methane is tracked over time. The results help us understand how climate affects emissions and provide tools that can be used worldwide to improve greenhouse gas monitoring and climate action planning.


Total:	0
HTML:	0
PDF:	0
XML:	0