The ICLIPS Climate Model (ICM) of Bruckner, Hooss, Füssel and Hasselmann (2003) is a reduced-form ('simple') climate model, implemented as time-discrete equality constraints in the GAMS optimization environment, that converts emissions of CO2, CH4, N2O, halocarbons, SF6 and SO2 into global-mean temperature, sea-level rise, and regionally and seasonally resolved fields of temperature, precipitation, cloud cover and humidity via empirical-orthogonal-function (EOF) scaling, with an explicitly nonlinear carbon cycle (seawater carbonate chemistry and CO2-fertilized biosphere uptake). It was built to make the Tolerable Windows Approach — an inverse, guardrail-based framework for climate-policy analysis — computationally tractable. The authors acknowledged a list of open issues for further work: dynamical inconsistency between heat and carbon advection (climate response calibrated to ECHAM3/LSG, carbon to HAMOCC); an incomplete terrestrial biosphere (no land-use change, biodiversity/productivity loss, net-primary-production saturation at high CO2, or warming-driven respiration); regional and seasonal patterns derived from a single GCM; the numerically solved nonlinear seawater-chemistry mapping lacking validation against a wider range of 3D model realizations; missing reduced-form aerosol and non-CO2 gas modules; unexplored climate-sensitivity parameter uncertainty inside the optimization; and unvalidated EOF scaling for variables with weaker climate-change signals. These acknowledged limitations define the objective: to determine to what extent they can be jointly removed from a single optimization-embeddable reduced-form climate model.

Background

The Tolerable Windows Approach (TWA), developed in the 1990s by the German Advisory Council on Global Change and the Potsdam Institute for Climate Impact Research (Petschel-Held et al., 1999), is an inverse approach to climate-policy analysis. Instead of evaluating a few exogenous emissions scenarios, one first fixes normative guardrails — limits on climate change (e.g., a maximum admissible global mean temperature increase) and limits on the socio-economically tolerable mitigation burden — and then computes the set of emissions pathways, the emissions corridor, that keeps the human–climate system within all guardrails (Bruckner, Petschel-Held, Leimbach and Tóth, 2003). Computing corridors requires solving very large numbers of intertemporal optimization problems, each embedding a full climate simulation, so the climate component must be a reduced-form model: fast enough to be executed thousands of times inside optimization software, yet complete enough to convert greenhouse-gas and aerosol emissions into the impact-relevant variables to which the guardrails refer (Bruckner et al., 2003; Füssel et al., 2003).

Two earlier strands supplied the ingredients. First, pulse-response studies (Joos et al., 1996) showed that the uptake of anthropogenic carbon by the ocean and biosphere simulated by complex three-dimensional models can be reproduced accurately by an impulse response function (IRF — the time-dependent response of a reservoir to a unit pulse input, whose convolution with an arbitrary emissions history approximates the full model's response). NICCS, the nonlinear impulse response model of the coupled carbon cycle–climate system (Hooss et al., 2001), extended this into the nonlinear regime by coupling the IRF of the HAMOCC ocean carbon-cycle model (an ocean-biogeochemistry model embedded in a three-dimensional ocean general circulation model; GCM denotes such a physically resolved atmosphere/ocean simulation model) to a differential box-model analogue and to the nonlinear inorganic carbonate chemistry of seawater. Second, pattern-scaling studies (Huntingford and Cox, 2000) showed that regional climate-change anomalies can be approximated as a fixed spatial pattern per degree of global mean warming; empirical orthogonal functions (EOFs — the spatial patterns obtained from a principal component analysis of a climate model's output fields, ordered by explained variance) supply such patterns, with the first principal component of near-surface temperature acting as the scalar scaling variable.

Within the ICLIPS project (Integrated Assessment of Climate Protection Strategies), Bruckner, Hooss, Füssel and Hasselmann (2003) assembled these ingredients into the ICLIPS Climate Model (ICM), implemented in GAMS (the General Algebraic Modeling System, a standard algebraic modeling and optimization environment) so that its differential equations become time-discrete equality constraints that nonlinear solvers can handle. The ICM converts emissions of CO2, CH4, N2O, halocarbons, SF6 and SO2 into concentrations and radiative forcings (non-CO2 modules adapted from the simple climate model MAGICC), evolves global mean temperature through a differential analogue of the ECHAM3/LSG coupled atmosphere–ocean GCM's climate IRF, treats nonlinear seawater chemistry and a logarithmically CO2-fertilized terrestrial biosphere in its carbon cycle, and scales regional and seasonal patterns of temperature, precipitation, cloud cover and humidity from EOFs of an 850-year ECHAM3/LSG transient warming experiment (the first EOF explains about 97% of the variance for annual mean temperature but only about 75%, 63% and 76% for precipitation, cloud cover and humidity). With this model the authors computed reachable climate domains — the set of temperature–concentration states attainable under bounded emissions — at a fixed climate sensitivity (equilibrium warming for doubled CO2) of 2.5 °C.

In their own discussion, the authors listed what the model still lacked: full dynamical consistency between heat and carbon advection (the carbon IRF comes from HAMOCC while the climate IRF and EOF patterns come from ECHAM3/LSG, so the same ocean does not advect both tracers); terrestrial-biosphere processes (land-use change and the associated loss of biological diversity and productivity, possible saturation of net primary production at high CO2, and accelerated respiratory carbon return under warming and hydrological feedbacks); regional and seasonal patterns from more than one GCM, with the corresponding uncertainty unquantified; wider validation of the numerically computed nonlinear seawater-chemistry mapping cs(c1)c_s(c_1) (the partition of anthropogenic carbon between the atmosphere and the oceanic mixed layer under carbonate equilibria, which has no closed-form solution); reduced-form aerosol and non-CO2 gas modules beyond the simplified MAGICC adaptations; exploration of climate-sensitivity parameter uncertainty inside the TWA optimization itself; and validation of the EOF-scaling approach for other impact-relevant variables and statistics where the climate-change signal is weaker. These items jointly define one scientific objective: a reduced-form climate model, embeddable in the optimization loop, whose global-mean and regional/seasonal outputs are dynamically consistent and validated against multiple full GCM/ESM (Earth system model) realizations, with quantified parameter and pattern uncertainty.

Problem Statement

Determine to what extent the following acknowledged limitations of the reduced-form climate model underpinning the Tolerable Windows Approach - the ICLIPS Climate Model or an equivalent optimization-embeddable reduced-form climate model - can be jointly removed under the in-loop tractability constraint defined below, as listed by the model's authors: (1) dynamical consistency of heat and carbon advection, deriving the climate response and the carbon-cycle response from the same parent ocean model or model ensemble, rather than combining a carbon impulse response from HAMOCC with a climate impulse response and EOF patterns from ECHAM3/LSG; (2) a terrestrial biosphere that represents land-use change and the associated losses of biological diversity and productivity, possible saturation (cut-off) of net primary production (the net rate at which vegetation absorbs carbon from the atmosphere) at high CO2, and accelerated respiratory carbon return due to warming and hydrological feedbacks; (3) regional and seasonal scaling patterns derived from several GCMs (at least two independent parent GCMs), with the uncertainty from the choice of parent GCM quantified rather than resting on a single ECHAM3/LSG experiment; (4) validation of the numerically computed nonlinear seawater-chemistry mapping cs(c1)c_s(c_1) and of the full model against a wider range of full 3D model realizations (at least two independent full 3D realizations; the absence of a closed form being an accepted property of the carbonate system, not itself the target); (5) incorporation of reduced-form aerosol and non-CO2 gas modules beyond the simplified MAGICC adaptations used in the ICM; (6) treatment of climate-sensitivity parameter uncertainty inside the TWA optimization itself; and (7) validation of the EOF-scaling approach for other impact-relevant variables and statistics where the climate-change signal is weaker than for temperature. The common direction of these items is a single reduced-form model, embeddable as equality constraints in the optimization loop, whose global-mean and regional/seasonal outputs are dynamically consistent, validated against multiple GCM/ESM realizations, and equipped with quantified parameter and pattern uncertainty. The in-loop tractability constraint is fixed by the task and carries one concrete bound that the solver does not choose: the climate model must remain expressible as time-discrete equality constraints in the TWA optimization loop, and one embedded model evaluation must run in well under a second, so that a corridor computation of at least a thousand evaluations finishes within minutes on commodity hardware. The seven items are weighed individually, not as a conjunction: (1) the thermal response and the carbon-cycle response derived from one and the same parent model or model ensemble; (2) a terrestrial biosphere whose land-use-change response is expressed in net primary production (the model-representable aspect of the associated loss of biological diversity and productivity), together with net-primary-production saturation at high CO2 and accelerated warming-driven respiratory return; (3) regional and seasonal patterns from at least two independent parent GCMs, with the parent-GCM uncertainty quantified; (4) the nonlinear seawater-chemistry mapping cs(c1)c_s(c_1) and the full model's global-mean and regional/seasonal fields validated against at least two independent full 3D realizations; (5) reduced-form aerosol and non-CO2 modules that go beyond the simplified MAGICC adaptations; (6) climate-sensitivity parameter uncertainty carried through the TWA optimization, so that the computed corridors depend on a sensitivity distribution rather than on a single fixed value; and (7) EOF scaling validated, for the weaker-signal variables of precipitation, cloud cover and humidity, against at least two independent full 3D realizations. The target is not patching the particular 2003 GAMS code base, which is no longer maintained.

References

  1. Bruckner, T., Hooss, G., Füssel, H.-M., Hasselmann, K. (2003). Climate System Modeling in the Framework of the Tolerable Windows Approach: The ICLIPS Climate Model. Climatic Change 56, 119-137. DOI: 10.1023/A:1021300924356. https://doi.org/10.1023/A:1021300924356
  2. Bruckner, T., Petschel-Held, G., Leimbach, M., Tóth, F. L. (2003). Methodological Aspects of the Tolerable Windows Approach. Climatic Change 56, 73-89. DOI: 10.1023/A:1021388429866. https://doi.org/10.1023/A:1021388429866
  3. Füssel, H.-M., Tóth, F. L., van Minnen, J. G., Kaspar, F. (2003). Climate Impact Response Functions as Impact Tools in the Tolerable Windows Approach. Climatic Change 56, 91-117. DOI: 10.1023/A:1021340513936. https://doi.org/10.1023/A:1021340513936
  4. Petschel-Held, G., Schellnhuber, H.-J., Bruckner, T., Tóth, F. L., Hasselmann, K. (1999). The Tolerable Windows Approach: Theoretical and Methodological Foundations. Climatic Change 41, 303-331. DOI: 10.1023/A:1005487123751. https://doi.org/10.1023/A:1005487123751
  5. Joos, F., Bruno, M., Fink, R., Siegenthaler, U., Stocker, T. F., Le Quéré, C., Sarmiento, J. L. (1996). An efficient and accurate representation of complex oceanic and biospheric models of anthropogenic carbon uptake. Tellus B 48, 397-417. DOI: 10.1034/j.1600-0889.1996.t01-2-00006.x. https://doi.org/10.1034/j.1600-0889.1996.t01-2-00006.x
  6. Hooss, G., Voss, R., Hasselmann, K., Maier-Reimer, E., Joos, F. (2001). A nonlinear impulse response model of the coupled carbon cycle-climate system (NICCS). Climate Dynamics 18, 189-202. DOI: 10.1007/s003820100170. https://doi.org/10.1007/s003820100170
  7. Huntingford, C., Cox, P. M. (2000). An analogue model to derive additional climate change scenarios from existing GCM simulations. Climate Dynamics 16, 575-586. DOI: 10.1007/s003820000067. https://doi.org/10.1007/s003820000067
  8. Kriegler, E., Bruckner, T. (2004). Sensitivity Analysis of Emissions Corridors for the 21st Century. Climatic Change 66, 345-387. DOI: 10.1023/B:CLIM.0000044615.75251.02. https://doi.org/10.1023/B:CLIM.0000044615.75251.02
  9. Zickfeld, K., Bruckner, T. (2008). Reducing the risk of Atlantic thermohaline circulation collapse: sensitivity analysis of emissions corridors. Climatic Change 91, 291-315. DOI: 10.1007/s10584-008-9467-0. https://doi.org/10.1007/s10584-008-9467-0
  10. Meinshausen, M., Raper, S. C. B., Wigley, T. M. L. (2011). Emulating coupled atmosphere-ocean and carbon cycle models with a simpler model, MAGICC6 - Part 1: Model description and calibration. Atmospheric Chemistry and Physics 11, 1417-1456. DOI: 10.5194/acp-11-1417-2011. https://doi.org/10.5194/acp-11-1417-2011
  11. Tanaka, K., O'Neill, B. C., Rokityanskiy, D., Obersteiner, M., Tol, R. S. J. (2009). Evaluating Global Warming Potentials with historical temperature. Climatic Change 96, 443-466. DOI: 10.1007/s10584-009-9566-6. https://doi.org/10.1007/s10584-009-9566-6
  12. Tanaka, K., O'Neill, B. C. (2018). The Paris Agreement zero-emissions goal is not always consistent with the 1.5 °C and 2 °C temperature targets. Nature Climate Change 8, 319-324. DOI: 10.1038/s41558-018-0097-x. https://doi.org/10.1038/s41558-018-0097-x
  13. Strassmann, K., Joos, F. (2018). The Bern Simple Climate Model (BernSCM) v1.0: an extensible and fully documented open-source re-implementation of the Bern reduced-form model for global carbon cycle-climate simulations. Geoscientific Model Development 11, 1887-1908. DOI: 10.5194/gmd-11-1887-2018. https://doi.org/10.5194/gmd-11-1887-2018
  14. Nicholls, Z. R. J., Meinshausen, M., Lewis, J., Gieseke, R., Dommenget, D., Dorheim, K., Fan, C.-S., Fuglestvedt, J. S., Gasser, T., Golüke, U., Goodwin, P., Hartin, C., Hope, A. P., Kriegler, E., Leach, N. J., Marchegiani, D., McBride, L. A., Quilcaille, Y., Rogelj, J., Salawitch, R. J., Samset, B. H., Sandstad, M., Shiklomanov, A. N., Skeie, R. B., Smith, C. J., Smith, S., Tanaka, K., Tsutsui, J., Xie, Z. (2020). Reduced Complexity Model Intercomparison Project Phase 1: introduction and evaluation of global-mean temperature response. Geoscientific Model Development 13, 5175-5190. DOI: 10.5194/gmd-13-5175-2020. https://doi.org/10.5194/gmd-13-5175-2020
  15. Tebaldi, C., Arblaster, J. M. (2014). Pattern scaling: Its strengths and limitations, and an update on the latest model simulations. Climatic Change 122, 459-471. DOI: 10.1007/s10584-013-1032-9. https://doi.org/10.1007/s10584-013-1032-9
  16. Herger, N., Sanderson, B. M., Knutti, R. (2015). Improved pattern scaling approaches for the use in climate impact studies. Geophysical Research Letters 42, 3486-3494. DOI: 10.1002/2015GL063569. https://doi.org/10.1002/2015GL063569
  17. Geogdzhayev, G., Souza, A. N., Flierl, G. R., Ferrari, R. (2026). An EOF-Based Emulator of Means and Covariances of Monthly Climate Fields. Earth System Dynamics 17, 235-263. DOI: 10.5194/esd-17-235-2026. https://doi.org/10.5194/esd-17-235-2026
  18. Steinacher, M., Joos, F., Stocker, T. F. (2013). Allowable carbon emissions lowered by multiple climate targets. Nature 499, 197-201. DOI: 10.1038/nature12269. https://doi.org/10.1038/nature12269
  19. Bossy, T., Ciais, P., Tanaka, K., Lecocq, F., Bousquet, P., Gasser, T. (2025). Spaces of anthropogenic CO2 emissions compatible with climate boundaries. Nature Climate Change 15, 1307-1314. DOI: 10.1038/s41558-025-02460-5. https://doi.org/10.1038/s41558-025-02460-5
  20. Xiong, W., Tanaka, K., Ciais, P., Johansson, D. J. A., Lehtveer, M. (2025). emIAM v1.0: an emulator for integrated assessment models using marginal abatement cost curves. Geoscientific Model Development 18, 1575-1612. DOI: 10.5194/gmd-18-1575-2025. https://doi.org/10.5194/gmd-18-1575-2025