2018 Update: Best Climate Model INMCM5 October 22, 2018 (follow link to latest info on INMCM5)

A previous analysis Temperatures According to Climate Models showed that only one of 42 CMIP5 models was close to hindcasting past temperature fluctuations. That model was INMCM4, which also projected an unalarming 1.4C warming to the end of the century, in contrast to the other models programmed for future warming five times the past.

In a recent comment thread, someone asked what has been done recently with that model, given that it appears to be “best of breed.” So I went looking and this post summarizes further work to produce a new, hopefully improved version by the modelers at the Institute of Numerical Mathematics of the Russian Academy of Sciences.

Earlier this year came this publication Simulation of the present-day climate with the climate model INMCM5 by E.M. Volodin et al. Excerpts below with my bolds.

In this paper we present the fifth generation of the INMCM climate model that is being developed at the Institute of Numerical Mathematics of the Russian Academy of Sciences (INMCM5). The most important changes with respect to the previous version (INMCM4) were made in the atmospheric component of the model. Its vertical resolution was increased to resolve the upper stratosphere and the lower mesosphere. A more sophisticated parameterization of condensation and cloudiness formation was introduced as well. An aerosol module was incorporated into the model. The upgraded oceanic component has a modified dynamical core optimized for better implementation on parallel computers and has two times higher resolution in both horizontal directions.

Analysis of the present-day climatology of the INMCM5 (based on the data of historical run for 1979–2005) shows moderate improvements in reproduction of basic circulation characteristics with respect to the previous version. Biases in the near-surface temperature and precipitation are slightly reduced compared with INMCM4 as  well as biases in oceanic temperature, salinity and sea surface height. The most notable improvement over INMCM4 is the capability of the new model to reproduce the equatorial stratospheric quasi-biannual oscillation and statistics of sudden stratospheric warmings.

 

The family of INMCM climate models, as most climate system models, consists of two main blocks: the atmosphere general circulation model, and the ocean general circulation model. The atmospheric part is based on the standard set of hydrothermodynamic equations with hydrostatic approximation written in advective form. The model prognostic variables are wind horizontal components, temperature, specific humidity and surface pressure.

Atmosphere Module

The INMCM5 borrows most of the atmospheric parameterizations from its previous version. One of the few notable changes is the new parameterization of clouds and large-scale condensation. In the INMCM5 cloud area and cloud water are computed prognostically according to Tiedtke (1993). That includes the formation of large-scale cloudiness as well as the formation of clouds in the atmospheric boundary layer and clouds of deep convection. Decrease of cloudiness due to mixing with unsaturated environment and precipitation formation are also taken into account. Evaporation of precipitation is implemented according to Kessler (1969).

In the INMCM5 the atmospheric model is complemented by the interactive aerosol block, which is absent in the INMCM4. Concentrations of coarse and fine sea salt, coarse and fine mineral dust, SO2, sulfate aerosol, hydrophilic and hydrophobic black and organic carbon are all calculated prognostically.

Ocean Module

The oceanic module of the INMCM5 uses generalized spherical coordinates. The model “South Pole” coincides with the geographical one, while the model “North Pole” is located in Siberia beyond the ocean area to avoid numerical problems near the pole. Vertical sigma-coordinate is used. The finite-difference equations are written using the Arakawa C-grid. The differential and finite-difference equations, as well as methods of solving them can be found in Zalesny etal. (2010).

The INMCM5 uses explicit schemes for advection, while the INMCM4 used schemes based on splitting upon coordinates. Also, the iterative method for solving linear shallow water equation systems is used in the INMCM5 rather than direct method used in the INMCM4. The two previous changes were made to improve model parallel scalability. The horizontal resolution of the ocean part of the INMCM5 is 0.5 × 0.25° in longitude and latitude (compared to the INMCM4’s 1 × 0.5°).

Both the INMCM4 and the INMCM5 have 40 levels in vertical. The parallel implementation of the ocean model can be found in (Terekhov etal. 2011). The oceanic block includes vertical mixing and isopycnal diffusion parameterizations (Zalesny et al. 2010). Sea ice dynamics and thermodynamics are parameterized according to Iakovlev (2009). Assumptions of elastic-viscous-plastic rheology and single ice thickness gradation are used. The time step in the oceanic block of the INMCM5 is 15 min.

Carbon Cycle Module

The climate model INMCM5 has а carbon cycle module (Volodin 2007), where atmospheric CO2 concentration, carbon in vegetation, soil and ocean are calculated. In soil, а single carbon pool is considered. In the ocean, the only prognostic variable in the carbon cycle is total inorganic carbon. Biological pump is prescribed. The model calculates methane emission from wetlands and has a simplified methane cycle (Volodin 2008). Parameterizations of some electrical phenomena, including calculation of ionospheric potential and flash intensity (Mareev and Volodin 2014), are also included in the model.

Surface Temperatures

When compared to the INMCM4 surface temperature climatology, the INMCM5 shows several improvements. Negative bias over continents is reduced mainly because of the increase in daily minimum temperature over land, which is achieved by tuning the surface flux parameterization. In addition, positive bias over southern Europe and eastern USA in summer typical for many climate models (Mueller and Seneviratne 2014) is almost absent in the INMCM5. A possible reason for this bias in many models is the shortage of soil water and suppressed evaporation leading to overestimation of the surface temperature. In the INMCM5 this problem was addressed by the increase of the minimum leaf resistance for some vegetation types.

Nevertheless, some problems migrate from one model version to the other: negative bias over most of the subtropical and tropical oceans, and positive bias over the Atlantic to the east of the USA and Canada. Root mean square (RMS) error of annual mean near surface temperature was reduced from 2.48 K in the INMCM4 to 1.85 K in the INMCM5.

Precipitation

In mid-latitudes, the positive precipitation bias over the ocean prevails in winter while negative bias occurs in summer. Compared to the INMCM4, the biases over the western Indian Ocean, Indonesia, the eastern tropical Pacific and the tropical Atlantic are reduced. A possible reason for this is the better reproduction of the tropical sea surface temperature (SST) in the INMCM5 due to the increase of the spatial resolution in the oceanic block, as well as the new condensation scheme. RMS annual mean model bias for precipitation is 1.35mm day−1 for the INMCM5 compared to 1.60mm day−1 for the INMCM4.

Cloud Radiation Forcing

Cloud radiation forcing (CRF) at the top of the atmosphere is one of the most important climate model characteristics, as errors in CRF frequently lead to an incorrect surface temperature.

In the high latitudes model errors in shortwave CRF are small. The model underestimates longwave CRF in the subtropics but overestimates it in the high latitudes. Errors in longwave CRF in the tropics tend to partially compensate errors in shortwave CRF. Both errors have positive sign near 60S leading to warm bias in the surface temperature here. As a result, we have some underestimation of the net CRF absolute value at almost all latitudes except the tropics. Additional experiments with tuned conversion of cloud water (ice) to precipitation (for upper cloudiness) showed that model bias in the net CRF could be reduced, but that the RMS bias for the surface temperature will increase in this case.

 

A table from another paper provides the climate parameters described by INMCM5.

Climate Parameters	Observations	INMCM3	INMCM4	INMCM5
Incoming solar radiation at TOA	341.3 [26]	341.7	341.8	341.4
Outgoing solar radiation at TOA	96–100 [26]	97.5 ± 0.1	96.2 ± 0.1	98.5 ± 0.2
Outgoing longwave radiation at TOA	236–242 [26]	240.8 ± 0.1	244.6 ± 0.1	241.6 ± 0.2
Solar radiation absorbed by surface	154–166 [26]	166.7 ± 0.2	166.7 ± 0.2	169.0 ± 0.3
Solar radiation reflected by surface	22–26 [26]	29.4 ± 0.1	30.6 ± 0.1	30.8 ± 0.1
Longwave radiation balance at surface	–54 to 58 [26]	–52.1 ± 0.1	–49.5 ± 0.1	–63.0 ± 0.2
Solar radiation reflected by atmosphere	74–78 [26]	68.1 ± 0.1	66.7 ± 0.1	67.8 ± 0.1
Solar radiation absorbed by atmosphere	74–91 [26]	77.4 ± 0.1	78.9 ± 0.1	81.9 ± 0.1
Direct hear flux from surface	15–25 [26]	27.6 ± 0.2	28.2 ± 0.2	18.8 ± 0.1
Latent heat flux from surface	70–85 [26]	86.3 ± 0.3	90.5 ± 0.3	86.1 ± 0.3
Cloud amount, %	64–75 [27]	64.2 ± 0.1	63.3 ± 0.1	69 ± 0.2
Solar radiation-cloud forcing at TOA	–47 [26]	–42.3 ± 0.1	–40.3 ± 0.1	–40.4 ± 0.1
Longwave radiation-cloud forcing at TOA	26 [26]	22.3 ± 0.1	21.2 ± 0.1	24.6 ± 0.1
Near-surface air temperature, °С	14.0 ± 0.2 [26]	13.0 ± 0.1	13.7 ± 0.1	13.8 ± 0.1
Precipitation, mm/day	2.5–2.8 [23]	2.97 ± 0.01	3.13 ± 0.01	2.97 ± 0.01
River water inflow to the World Ocean,10^3 km^3/year	29–40 [28]	21.6 ± 0.1	31.8 ± 0.1	40.0 ± 0.3
Snow coverage in Feb., mil. Km^2	46 ± 2 [29]	37.6 ± 1.8	39.9 ± 1.5	39.4 ± 1.5
Permafrost area, mil. Km^2	10.7–22.8 [30]	8.2 ± 0.6	16.1 ± 0.4	5.0 ± 0.5
Land area prone to seasonal freezing in NH, mil. Km^2	54.4 ± 0.7 [31]	46.1 ± 1.1	48.3 ± 1.1	51.6 ± 1.0
Sea ice area in NH in March, mil. Km^2	13.9 ± 0.4 [32]	12.9 ± 0.3	14.4 ± 0.3	14.5 ± 0.3
Sea ice area in NH in Sept., mil. Km^2	5.3 ± 0.6 [32]	4.5 ± 0.5	4.5 ± 0.5	6.1 ± 0.5

Heat flux units are given in W/m^2; the other units are given with the title of corresponding parameter. Where possible, ± shows standard deviation for annual mean value.  Source: Simulation of Modern Climate with the New Version Of the INM RAS Climate Model (Bracketed numbers refer to sources for observations)

Ocean Temperature and Salinity

The model biases in potential temperature and salinity averaged over longitude with respect to WOA09 (Antonov et al. 2010) are shown in Fig.12. Positive bias in the Southern Ocean penetrates from the surface downward for up to 300 m, while negative bias in the tropics can be seen even in the 100–1000 m layer.

Nevertheless, zonal mean temperature error at any level from the surface to the bottom is small. This was not the case for the INMCM4, where one could see negative temperature bias up to 2–3 K from 1.5 km to the bottom nearly al all latitudes, and 2–3 K positive bias at levels of 700–1000 m. The reason for this improvement is the introduction of a higher background coefficient for vertical diffusion at high depth (3000 m and higher) than at intermediate depth (300–500m). Positive temperature bias at 45–65 N at all depths could probably be explained by shortcomings in the representation of deep convection [similar errors can be seen for most of the CMIP5 models (Flato etal. 2013, their Fig.9.13)].

Another feature common for many present day climate models (and for the INMCM5 as well) is negative bias in southern tropical ocean salinity from the surface to 500 m. It can be explained by overestimation of precipitation at the southern branch of the Inter Tropical Convergence zone. Meridional heat flux in the ocean (Fig.13) is not far from available estimates (Trenberth and Caron 2001). It looks similar to the one for the INMCM4, but maximum of northward transport in the Atlantic in the INMCM5 is about 0.1–0.2 × 1015 W higher than the one in the INMCM4, probably, because of the increased horizontal resolution in the oceanic block.

Sea Ice

In the Arctic, the model sea ice area is just slightly overestimated. Overestimation of the Arctic sea ice area is connected with negative bias in the surface temperature. In the same time, connection of the sea ice area error with the positive salinity bias is not evident because ice formation is almost compensated by ice melting, and the total salinity source for these pair of processes is not large. The amplitude and phase of the sea ice annual cycle are reproduced correctly by the model. In the Antarctic, sea ice area is underestimated by a factor of 1.5 in all seasons, apparently due to the positive temperature bias. Note that the correct simulation of sea ice area dynamics in both hemispheres simultaneously is a difficult task for climate modeling.

The analysis of the model time series of the SST anomalies shows that the El Niño event frequency is approximately the same in the model and data, but the model El Niños happen too regularly. Atmospheric response to the El Niño vents is also underestimated in the model by a factor of 1.5 with respect to the reanalysis data.

Conclusion

Based on the CMIP5 model INMCM4 the next version of the Institute of Numerical Mathematics RAS climate model was developed (INMCM5). The most important changes include new parameterizations of large scale condensation (cloud fraction and cloud water are now the prognostic variables), and increased vertical resolution in the atmosphere (73 vertical levels instead of 21, top model level raised from 30 to 60 km). In the oceanic block, horizontal resolution was increased by a factor of 2 in both directions.

The climate model was supplemented by the aerosol block. The model got a new parallel code with improved computational efficiency and scalability. With the new version of climate model we performed a test model run (80 years) to simulate the present-day Earth climate. The model mean state was compared with the available datasets. The structures of the surface temperature and precipitation biases in the INMCM5 are typical for the present climate models. Nevertheless, the RMS error in surface temperature, precipitation as well as zonal mean temperature and zonal wind are reduced in the INMCM5 with respect to its previous version, the INMCM4.

The model is capable of reproducing equatorial stratospheric QBO and SSWs.The model biases for the sea surface height and surface salinity are reduced in the new version as well, probably due to increasing spatial resolution in the oceanic block. Bias in ocean potential temperature at depths below 700 m in the INMCM5 is also reduced with respect to the one in the INMCM4. This is likely because of the tuning background vertical diffusion coefficient.

Model sea ice area is reproduced well enough in the Arctic, but is underestimated in the Antarctic (as a result of the overestimated surface temperature). RMS error in the surface salinity is reduced almost everywhere compared to the previous model except the Arctic (where the positive bias becomes larger). As a final remark one can conclude that the INMCM5 is substantially better in almost all aspects than its previous version and we plan to use this model as a core component for the coming CMIP6 experiment.

Summary

One the one hand, this model example shows that the intent is simple: To represent dynamically the energy balance of our planetary climate system.  On the other hand, the model description shows how many parameters are involved, and the complexity of processes interacting.  The attempt to simulate operations of the climate system is a monumental task with many outstanding challenges, and this latest version is another step in an iterative development.

Note:  Regarding the influence of rising CO2 on the energy balance.  Global warming advocates estimate a CO2 perturbation of 4 W/m^2.  In the climate parameters table above, observations of the radiation fluxes have a 2 W/m^2 error range at best, and in several cases are observed in ranges of 10 to 15 W/m^2.

We do not yet have access to the time series temperature outputs from INMCM5 to compare with observations or with other CMIP6 models.  Presumably that will happen in the future.

A previous post Chameleon Climate Models described the general issue of whether a model belongs on the bookshelf (theoretically useful) or whether it passes real world filters of relevance, thus qualifying as useful for policy considerations.

Following an interesting discussion on her blog, Dr. Judith Curry has written an important essay on the usefulness and limitations of climate models.

The paper was developed to respond to a request from a group of lawyers wondering how to regard claims based upon climate model outputs. The document is entitled Climate Models (here) and is a great informative read for anyone. Some excerpts that struck me:

Climate model development has followed a pathway mostly driven by scientific curiosity and computational limitations. GCMs were originally designed as a tool to help understand how the climate system works. GCMs are used by researchers to represent aspects of climate that are extremely difficult to observe, experiment with theories in a new way by enabling hitherto infeasible calculations, understand a complex system of equations that would otherwise be impenetrable, and explore the climate system to identify unexpected outcomes. As such, GCMs are an important element of climate research.

Climate models are useful tools for conducting scientific research to understand the climate system. However, the above points support the conclusion that current GCM climate models are not fit for the purpose of attributing the causes of 20th century warming or for predicting global or regional climate change on timescales of decades to centuries, with any high level of confidence. By extension, GCMs are not fit for the purpose of justifying political policies to fundamentally alter world social, economic and energy systems. It is this application of climate model results that fuels the vociferousness of the debate surrounding climate models.

The actual equations used in the GCM computer codes are only approximations of the physical processes that occur in the climate system. While some of these approximations are highly accurate, others are unavoidably crude. This is because the real processes they represent are either poorly understood or too complex to include in the model given the constraints of the computer system. Of the processes that are most important for climate change, parameterizations related to clouds and precipitation remain the most challenging, and are the greatest source of disagreement among different GCMs.

There are literally thousands of different choices made in the construction of a climate model (e.g. resolution, complexity of the submodels, parameterizations). Each different set of choices produces a different model having different sensitivities. Further, different modeling groups have different focal interests, e.g. long paleoclimate simulations, details of ocean circulations, nuances of the interactions between aerosol particles and clouds, the carbon cycle. These different interests focus their limited computational resources on a particular aspect of simulating the climate system, at the expense of others.

Overview of the structure of a state-of-the-art climate model. See Climate Models Explained by R.G. Brown

Human-caused warming depends not only on how much CO2 is added to the atmosphere, but also on how ‘sensitive’ the climate is to the increased CO2. Climate sensitivity is defined as the global surface warming that occurs when the concentration of carbon dioxide in the atmosphere doubles. If climate sensitivity is high, then we can expect substantial warming in the coming century as emissions continue to increase. If climate sensitivity is low, then future warming will be substantially lower.

In GCMs, the equilibrium climate sensitivity is an ‘emergent property’ that is not directly calibrated or tuned. While there has been some narrowing of the range of modeled climate sensitivities over time, models still can be made to yield a wide range of sensitivities by altering model parameterizations. Model versions can be rejected or not, subject to the modelers’ own preconceptions, expectations and biases of the outcome of equilibrium climate sensitivity calculation.

Further, the discrepancy between observational and climate model-based estimates of climate sensitivity is substantial and of significant importance to policymakers. Equilibrium climate sensitivity, and the level of uncertainty in its value, is a key input into the economic models that drive cost-benefit analyses and estimates of the social cost of carbon.

Variations in climate can be caused by external forcing, such as solar variations, volcanic eruptions or changes in atmospheric composition such as an increase in CO2. Climate can also change owing to internal processes within the climate system (internal variability). The bestknown example of internal climate variability is El Nino/La Nina. Modes of decadal to centennial to millennial internal variability arise from the slow circulations in the oceans. As such, the ocean serves as a ‘fly wheel’ on the climate system, storing and releasing heat on long timescales and acting to stabilize the climate. As a result of the time lags and storage of heat in the ocean, the climate system is never in equilibrium.

The combination of uncertainty in the transient climate response (sensitivity) and the uncertainties in the magnitude and phasing of the major modes in natural internal variability preclude an unambiguous separation of externally forced climate variations from natural internal climate variability. If the climate sensitivity is on the low end of the range of estimates, and natural internal variability is on the strong side of the distribution of climate models, different conclusions are drawn about the relative importance of human causes to the 20th century warming.

Anthropogenic (human-caused) climate change is a theory in which the basic mechanism is well understood, but whose potential magnitude is highly uncertain. What does the preceding analysis imply for IPCC’s ‘extremely likely’ attribution of anthropogenically caused warming since 1950? Climate models infer that all of the warming since 1950 can be attributed to humans. However, there have been large magnitude variations in global/hemispheric climate on timescales of 30 years, which are the same duration as the late 20th century warming. The IPCC does not have convincing explanations for previous 30 year periods in the 20th century, notably the warming 1910-1945 and the grand hiatus 1945-1975. Further, there is a secular warming trend at least since 1800 (and possibly as long as 400 years) that cannot be explained by CO2, and is only partly explained by volcanic eruptions.

Summary

There is growing evidence that climate models are running too hot and that climate sensitivity to CO2 is on the lower end of the range provided by the IPCC. Nevertheless, these lower values of climate sensitivity are not accounted for in IPCC climate model projections of temperature at the end of the 21st century or in estimates of the impact on temperatures of reducing CO2 emissions.

The climate modeling community has been focused on the response of the climate to increased human caused emissions, and the policy community accepts (either explicitly or implicitly) the results of the 21st century GCM simulations as actual predictions. Hence we don’t have a good understanding of the relative climate impacts of the above (natural factors) or their potential impacts on the evolution of the 21st century climate.

Footnote:

There are a series of posts here which apply reality filters to attest climate models.  The first was Temperatures According to Climate Models where both hindcasting and forecasting were seen to be flawed.

Others in the Series are:

Sea Level Rise: Just the Facts

Data vs. Models #1: Arctic Warming

Data vs. Models #2: Droughts and Floods

Data vs. Models #3: Disasters

Data vs. Models #4: Climates Changing

Climate Medicine

Climates Don’t Start Wars, People Do