Forecasting Precipitation and Temperature in the Qareh-Sou Watershed: Emphasizing Model Uncertainties

Document Type : Research Article

Authors

• Golaleh Ghaffari ¹
• Ali Salajegh ²

¹ Assistant Professor of Soil Conservation and Watershed Management, Research and Education Center for Agriculture and Natural Resources of Kermanshah Province, Kermanshah, Iran

² Professor of Faculty of Natural Resources, University of Tehran, Tehran, Iran

10.22067/geoeh.2024.86195.1450

Abstract

As part of the discussion on climate change, one of the most significant sources of uncertainty is the use of different Atmosphere-Ocean General Circulation Models (AOGCMs), which produce varying outputs for climatic variables. The general consensus is that employing multiple AOGCM models or ensemble methods (Ensemble Projection, EP) for weather forecasting—to account for uncertainty—is a strategy to mitigate the uncertainties associated with AOGCMs. These uncertainties arise from structural differences between global climate models and variations in initial conditions.
For this study, synoptic station data from the Kermanshah Meteorological Organization (spanning January 1, 1993, to December 31, 2016) and downscaled data from 21 AOGCM models (obtained from NASA’s NEX-GDDP dataset) were analyzed. The goal was to assess AOGCM model uncertainty for the past period (1976–2005) and the future period (2020–2049) under the RCP4.5 scenario, as well as to reduce uncertainty using ensemble methods.
The results of the model comparison revealed that the MRI-CGCM3, MPI-ESM-LR, BNU-ESM, ACCESS1-0, MIROC-ESM, MIROC-ESM-CHEM, and MPI-ESM-MR models performed better in simulations. As expected, the models with the lowest errors received the highest weights, indicating that they were the most reliable for simulating precipitation, maximum temperature, and minimum temperature in the past period. These models can thus be considered the most suitable for future predictions, with the least uncertainty in temperature and precipitation simulations.
An examination of statistical coefficients from different ensemble methods showed that the Modified Ensemble Projection (MEP) method provided the best estimation, with a coefficient of determination (R²) of 0.95 and an efficiency coefficient (ME) of 0.92, compared to baseline data from the Kermanshah synoptic station. Consequently, this method was identified as the optimal ensemble approach for AOGCM models.
Extended Abstract
Introduction
 Climate change, driven by global warming, poses an existential threat to both natural and human systems, necessitating accurate and reliable projections of future climate variations. While extensive research has examined the impacts of climate change on various subsystems and proposed adaptation and mitigation strategies, many studies have overlooked the inherent uncertainties in climate modeling. Limiting analyses to selected scenarios from Atmosphere-Ocean General Circulation Models (AOGCMs) and neglecting uncertainty analysis reduces the credibility and certainty of the final results. Indeed, uncertainties arising from structural and parametric diversity in models, downscaling processes, and impact assessment models require rigorous investigation and quantification.
A comprehensive assessment of climate change impacts requires the identification and analysis of three primary categories of uncertainty:

Uncertainties associated with the structure and parameters of AOGCMs;
Uncertainties arising from statistical and dynamical downscaling methods at regional scales; and
Parametric and structural uncertainties in impact assessment models.

AOGCMs, due to their diversity in structure and parameters, produce varying outputs for climatic variables, representing a significant source of uncertainty in climate projections. Incorporating uncertainties into climate change impact assessments leads to various plausible future scenarios that can inform decision-making and adaptation planning. Given computational and data limitations, a comprehensive examination of all uncertainty sources in every study is infeasible. However, studies have shown that certain uncertainties, such as those related to downscaling methods, significantly impact the final results and must be considered in assessing climate change impacts on river flows. Employing multi-model ensemble approaches and ensemble prediction methods to quantify and reduce uncertainties arising from AOGCMs is an effective strategy adopted in this research.
Material and Method
The Qarasu watershed, located in western Iran within Kermanshah province, served as the focal point of this study. Observational data spanning 1976–2005 were obtained from the Kermanshah synoptic station, courtesy of the Iran Meteorological Organization. Complementing this, downscaled data from 21 Global Circulation Models (GCMs) were acquired from the NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) dataset. This dataset encompassed both the historical (1976–2005) and future (2020–2049) periods under the Representative Concentration Pathway 4.5 (RCP4.5) scenario, selected for its intermediate climate change trajectory.
Following meticulous data validation and preprocessing, the uncertainty inherent in the GCMs was rigorously assessed. To evaluate GCM performance, the coefficient of determination (R²) and Nash-Sutcliffe Efficiency (NSE) were computed, facilitating a comparative analysis between simulated temperature and precipitation under RCP4.5 and observed values from the Kermanshah synoptic station.
To address and mitigate uncertainties in climate projections, this research employed a suite of ensemble methods, including ensemble prediction (EP), multi-model ensembles (MEP), and weighted multi-model ensembles (MEPWi), rather than relying solely on individual model outputs. The fundamental premise underpinning these methods is that models demonstrating superior skill in replicating historical climate patterns are anticipated to maintain their relative accuracy in future projections, thereby identifying them as optimal candidates. Consequently, the weight assigned to each model within the ensemble is directly proportional to its historical performance. Finally, the accuracy of the model simulations relative to observational data was evaluated using R² and NSE.
Results and Discussion
The findings indicated that MRI-CGCM3, MPI-ESM-LR, BNU-ESM, ACCESS1-0, MIROC-ESM, MIROC-ESM-CHEM, and MPI-ESM-MR exhibited robust performance in simulating monthly precipitation patterns. Concurrently, ACCESS1-0, CNRM-CM5, MIROC-ESM, MIROC-ESM-CHEM, MPI-ESM-LR, MPI-ESM-MR, MRI-CGCM3, and BNU-ESM demonstrated heightened accuracy in replicating temperature regimes. Notably, MRI-CGCM3, MPI-ESM-MR, and MIROC-ESM-CHEM were assigned the highest weights and exhibited the lowest uncertainty in their simulations of monthly precipitation, maximum temperature, and minimum temperature, signifying their superior fidelity and minimal deviation.
Analysis of statistical metrics from the ensemble methods revealed that the multi-model ensemble prediction (MEP) approach, characterized by an R² of 0.95 and NSE of 0.92, provided the most congruent estimates compared to baseline data from the Kermanshah synoptic station. Consequently, the MEP method was adjudicated as the optimal ensemble prediction paradigm for GCMs in this study.
Scrutiny of mean monthly and annual fluctuations under the MEP framework projected that:

Monthly and annual precipitation are anticipated to change by 1.9% and 22.7%, respectively, in the future period.
Mean monthly and annual temperature increments are projected to be 1.89°C and 1.88°C, respectively.

Conclusions
This study examined and compared various climate modeling methodologies to mitigate uncertainty in climate projections. Our findings reveal that no single climate model accurately predicts all climatic parameters within a given region. Optimal projections for temperature and precipitation necessitate the utilization of multiple models, underscoring the importance of multi-model ensemble techniques. We evaluated three ensemble methods: Model Ensemble with Weighting (MEPWi), Model Ensemble Projection (MEP), and Ensemble Prediction (EP).
The results indicated that projected values from these methods were relatively consistent, with minimal discernible differences. However, the MEP method yielded the most precise estimates for temperature and precipitation, establishing it as the superior technique for reducing uncertainty in climate projections. This research emphasizes that leveraging diverse climate models significantly enhances projection accuracy and reduces uncertainties. Relying solely on a single GCM is insufficient for formulating robust strategies to mitigate climate change impacts. Our results align with prior research demonstrating the efficacy of multi-model ensembles in improving predictive accuracy.
In summary, this study demonstrates that applying climate model ensemble techniques, particularly the MEP method, substantially improves the reliability of climate projections. This enhanced precision is crucial for enabling policymakers and planners to make informed decisions aimed at mitigating climate change impacts. Future studies could explore the performance of these ensemble methods under different climate scenarios and in other regions.
 

Main Subjects

• Climate change

---

References

Andarzian, ‌B., Bannayan, M., Steduto, P., Mazraeh, H., Barati, M. E., Barati, M. A., & Rahnama, A. (2011). Validation and testing of the AquaCrop model under full and deficit irrigated wheat production in Iran.  Agricultural Water Management, 100(1), 1-8. https://doi.org/10.1016/j.agwat.2011.08.023

Ashraf, B., Alizadeh, A., Mousavi Baygi, M., & Bannayan Awal, M. (2014). Verification of Temperature and Precipitation Simulated Data by Individual and Ensemble Performance of Five AOGCM Models for North East of Iran. Water and Soil, 28(2), 253-266. [In Persian] https://doi.org/10.22067/jsw.v0i0.38011

Cerqueira, V., & Torgo, L. (2023). Multi-output Ensembles for Multi-step Forecasting. ArXiv Preprint ArXiv:2306.14563. https://doi.org/10.48550/arXiv.2306.14563

Chen, H., Zhang, R., Liu, F., Shao, C., Liu, F., Li, W., ... & Lou, B. (2023). The chromosome-level genome of Cherax quadricarinatus. Scientific Data, 10(1), 215. https://doi.org/10.1038/s41597-023-02124-z

Chilkoti, V., Bolisetti, T., & Balachandar, R. (2017). Climate change impact assessment on hydropower generation using multi-model climate ensemble. Renewable Energy, 109, 510-517. https://doi.org/10.1016/j.renene.2017.02.041

Effron, B., & Tibshirani, R.J. (1993). An introduction to the bootstrap. New York: Chapman and Hall/CRC. https://doi.org/10.1201/9780429246593

Fowler, H. J., & Ekström, M. (2009). Multi‐model ensemble estimates of climate change impacts on UK seasonal precipitation extremes. International Journal of Climatology: A Journal of the Royal Meteorological Society, 29(3), 385-416.  https://doi.org/10.1002/joc.1827

Fu, C., Wang, S., Xiong, Z., Gutowski, W. J., Lee, D. K., McGregor, J. L., ... & Suh, M. S. (2005). Regional climate model intercomparison project for Asia. Bulletin of the American Meteorological Society, 86(2), 257-266. https://doi.org/10.1175/BAMS-86-2-257

Gao, X., Sokolov, A., & Schlosser, C.A .(2023). A large ensemble global dataset for climate impact assessments. Scientific Data, 10, 801. https://doi.org/10.1038/s41597-023-02708-9

Heidari, M. & Khazaei, M. R. (2018). Climate change impact assessment on extreme daily rainfalls in Kermanshah. Journal of Water and Soil Resources Conservation, (7)2,27-39. [In Persian] https://journals.srbiau.ac.ir/article_11544.html?lang=en  

Jahangir, M. H., & Azimi, S. M. E. (2022). Evaluating the performance of artificial intelligence models for temperature downscaling (Study area: Ardabil province). Environmental Sciences, 20(4), 243-258. [In Persian] https://doi.org/10.48308/envs.2022.35101

Khazaei, M.R. & Khazaei, H. (2018). Scenarios in climate change impact assessment on monthly stream-flow of Karun basin. Journal of Environmental Sciences and Technology, 20(1),29-40. [In Persian]

Lane, M. E., Kirshen, P. H., & Vogel, R. M. (1999). Indicators of impacts of global climate change on US water resources. Journal of Water Resources Planning and Management, 125(4), 194-204. https://doi.org/10.1061/(ASCE)0733-9496(1999)125:4(194)

Mitchell, T.D. (2003). Pattern scaling: an examination of the accuracy of the technique for describing future climates. Climatic Change, 60(3), 217-242. https://doi.org/10.1023/A:1026035305597

Pinto, I., de Perez, E. C., Jaime, C., Wolski, P., van Aardenne, L., Jjemba, E., ... & Tall, A. (2023). Climate change projections from a multi-model ensemble of CORDEX and CMIPs over Angola. Environmental Research: Climate, 2(3), 035007. https://doi.org/10.1088/2752-5295/ace210

Semenov, M. A., & Stratonovitch, P. (2010). Use of multi-model ensembles from global climate models for assessment of climate change impacts. Climate Research, 41(1), 1-14. https://doi.org/10.3354/cr00836

Whetton,P., Hennessy, K., Bates, B., & Kent, D.(2010). Regional Projections and Model Evaluation: Potential Benefits of ‘Representative Future Regional Climates’. In Stocker, T., Dahe, Q., Plattner, G. K., Tignor, M., & Midgley, P., IPCC expert meeting on assessing and combining multi model climate projections. Paper presented at Proceedings of the National Center for Atmospheric Research, Boulder Colorado, USA(pp105-107). Switzerland: IPCC Working Group.https://www.ipcc.ch/publication/ipcc-expert-meeting-on-assessing-and-combining-multi-model-climate-projections/

Wilby, R. L., & Harris, I. (2006). A framework for assessing uncertainties in climate change impacts: Low‐flow scenarios for the River Thames, UK. Water Resources Research, 42(2),1-10. https://doi.org/10.1029/2005WR004065

Wilby, R. L., Tomlinson, O. J., & Dawson, C. W. (2003). Multi-site simulation of precipitation by conditional resampling. Climate Research, 23(3), 183–194. http://www.jstor.org/stable/24868347

Xu, K., Xu, B., Ju, J., Wu, C., Dai, H., & Hu, B. X. (2019). Projection and uncertainty of precipitation extremes in the CMIP5 multimodel ensembles over nine major basins in China. Atmospheric Research, 226, 122-137. https://doi.org/10.1016/j.atmosres.2019.04.018