Assessing the application of random forest (RF) to predict water-table (WT) in selected Irish peatlands: Challenges and opportunities with upscaling the model to national scale

Peatlands are a well-known major global terrestrial carbon (C) sink. Most of Irish peatlands have been anthropogenically altered in past, such as by drainage for forestry, agriculture, and peat extraction. Recent restoration efforts highlight the need to better understand peatland C-dynamics and improve methods for reporting CO₂ fluxes. Water-table (WT) depth is a key driver of CO₂ flux exchange under different land-use types [1], making its inclusion in predictive models essential. Previous work by Premrov et al. [2], [3] assessed the application of random forest (RF) model to predict WT [2] at eight Irish peatland sites [4], using on-site measurements [4] and gridded meteorological data [5], with further details explained in Premrov, et al (2025) [2]. Findings reported in Premrov et al. (2025) [2] showed a relatively good RF- model performance at site-scale (R² = 0.78). New work explores challenges and opportunities in upscaling the RF model from site- to national-scale. The ‘variable importance’ was assessed in order to reduce dimensionality (i.e. excluding less important variables), and to enable the upscaling. This resulted in a simplified RF model using only selected variables with the highest importance, to improve computational efficiency for upscaling model applications. While the simplified RF model showed slightly lower, but acceptable performance at site-scale compared to full RF model (including all variables) [2], the attempts to apply it at the national scale revealed challenges, and highlighted the need for further improvements. The study discusses the challenges and opportunities in upscaling the RF model to enhance the robustness of RF-based WT predictions at national scale.

 

Acknowledgements

The authors are grateful to the Irish Environmental Protection Agency (EPA) for funding projects CO2PEAT (2022-CE-1100) and AUGER (2015-CCRP-MS.30) [EPA Research Programmes 2021- 2030 and 2014–2020], and to University of Limerick funding.

 

References

[1] Tiemeyer, B., et al., 2020. A new methodology for organic soils in national greenhouse gas inventories: Data synthesis, derivation and application,Ecological Indicators, Vol. 109, 105838,  https://doi.org/10.1016/j.ecolind.2019.105838.

[2] Premrov, A., et.al., 2025. Assessing the application of random forest (RF) to predict water-table (WT) in selected Irish peatlands, EGU25-5122, https://doi.org/10.5194/egusphere-egu25-5122, 2025. 

[3] Premrov, A., et.al, 2023. Insights into the CO2PEAT project: Improving methodologies for reporting and verifying terrestrial CO2 removals and emissions from Irish peatlands. IGRM2023, Belfast, UK.  https://www.researchgate.net/publication/369061601_Insights_into_the_CO2PEAT_project_Improving_methodologies_for_reporting_and_verifying_terrestrial_CO2_removals_and_emissions_from_Irish_peatlands.

[4] Renou-Wilson, F., et. al, 2022. Peatland Properties Influencing Greenhouse Gas  Emissions and Removal (AUGER Project) (2015-CCRP-MS.30), EPA Research Report, Irish Environmental Protection Agency (EPA). https://www.epa.ie/publications/research/climate-change/Research_Report_401.pdf.

[5] Copernicus Climate Change Service, Climate Data Store, (2020): E-OBS daily gridded meteorological data for Europe from 1950 to present derived from in-situ observations. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). https://doi.org/10.24381/cds.151d3ec6.

[6] Kuhn, M. 2008. Building Predictive Models in R Using the caret Package. Journal of Statistical Software, 28(5), 1–26. https://doi.org/10.18637/jss.v028.i05.