Impact of parametric uncertainty on simulated climate extremes and attribution studies

<p>The attribution of extreme weather events, such as heavy rainfall, to anthropogenic influence typically involves the analysis of their probability in simulations of climate, such as those conducted in the C20C+ Detection and Attribution Project. The climate models used however, such as the Community Atmosphere Model (CAM), employ approximate physics that gives rise to “parameter uncertainty”—uncertainty about the most accurate or optimal values of numerical parameters within the model. Parameterisations for convective processes, for example, are well known to be influential in the simulation of precipitation extremes.</p><p>In the context of extreme event attribution, we investigate the importance of components of parameterisations—through their associated tuning parameters—relating to deep and shallow convection, and cloud and aerosol microphysics in CAM. We present results from the analysis of a large perturbed physics ensemble experiment (~12,000 years of simulation, ~1 degree horizontal resolution) designed to explore extremes in both the observed world and pre-industrial conditions. Using surrogate models based upon Gaussian processes fitted marginally to both regional and grid cell output, we have computed sensitivity measures associated with the physics parameters, for precipitation and temperature extremes and their respective “risk ratios”.</p><p>Our results reveal the high geospatial variability in averages and extremes of output variables arising from physics perturbations, and how this contrasts with low variability in estimates of risk ratios based upon the same variables. We conclude that for CAM, variability induced by perturbed physics is typically consistent across warming scenarios, and unlikely to be a significant source of uncertainty in extreme event attribution studies. However, we caution that this may not be the case in regions where relevant parameterisations are strongly active.</p>

Improving climate model accuracy by exploring parameter space with an O(10⁵) member ensemble and emulator

Understanding the unfolding challenges of climate change relies on climate models, many of which have large summer warm and dry biases over Northern Hemisphere continental mid-latitudes. This work, using the example of the model used in the updated version of the weather@home distributed climate model framework, shows the potential for improving climate model simulations through a multi-phased parameter refinement approach, particularly over northwestern United States(NWUS). Each phase consists of 1) creating a perturbed physics ensemble with the coupled global – regional atmospheric model, 2) building statistical emulators that estimate climate metrics as functions of parameter values, 3) and using the emulators to further refine the parameter space. The refinement process includes sensitivity analyses to identify the most influential parameters for various model output metrics; results are then used to cull parameters with little influence. Three phases of this iterative process are carried out before the results are considered to be satisfactory; that is, a handful of parameter sets are identified that meet acceptable bias reduction criteria. Results not only indicate that 74 % of the NWUS regional warm biases can be reduced by refining global atmospheric parameters that control convection and hydrometeor transport, and land surface parameters that affect plant photosynthesis, transpiration and evaporation, but also suggest that this iterative approach to perturbed physics has an important role to play in the evolution of physical parameterizations.

Finding Ocean States That Are Consistent with Observations from a Perturbed Physics Parameter Ensemble

A very large ensemble is used to identify subgrid-scale parameter settings for the HadCM3 model that are capable of best simulating the ocean state over the recent past (1980–2010). A simple particle filtering technique based upon the agreement of basin mean sea surface temperature (SST) and upper 700-m ocean heat content with EN3 observations is applied to an existing perturbed physics ensemble with initial conditions perturbations. A single set of subgrid-scale parameter values was identified from the wide range of initial parameter sets that gave the best agreement with ocean observations for the period studied. The parameter set, different from the standard model parameters, has a transient climate response of 1.68 K. The selected parameter set shows an improved agreement with EN3 decadal-mean SST patterns and the Atlantic meridional overturning circulation (AMOC) at 26°N as measured by the Rapid Climate Change (RAPID) array. Particle filtering techniques as demonstrated here could have a useful role in improving the starting point for traditional model-tuning exercises in coupled climate models.

Analysis of within and between-GCM uncertainties of runoff projections in Mediterranean-like catchments

Streamflow reductions have been reported in mid-latitude Southern Hemisphere (SH) catchments, in particular in the southwest of Western Australia (SWA) and in central Chile (CC), following decreases in precipitation since the mid-1970s. Although projections from Global Climate Models (GCMs) indicate the observed trends are expected to continue during the rest of the 21st century, they are affected by large uncertainties that challenge informed decision making. Quantification and comparison of uncertainties in runoff projections for the period 2050-2080 relative to 1970-2000, driven by an ensemble of a single GCM with perturbed physics (CPDN) and a multi-model ensemble of different GCMs (CMIP5), were used to account for what we term “within-GCM” and “between-GCM” uncertainty in SWA catchments. Between GCM uncertainty of runoff projections was also quantified in CC catchments. Within and between-GCM uncertainties were found to be very similar (∼55 per cent) in SWA catchments. Between-GCM uncertainty for runoff projections in CC catchments is smaller than in SWA. On average, uncertainty of about 51 per cent, under RCP8.5 scenario, was simulated for the period 2050-2080 compared to 1970-2000. For CC catchments a dichotomy was observed in runoff projections under the RCP4.5 scenario, which according to our preliminary analysis might relate to how ozone is specified within different GCMs. We conclude that the number of models sampled by the CMIP5 ensemble, which includes multiple model runs from some GCMs, provides some insight into within-GCM uncertainties. Furthermore, since CMIP5 model runs report values for all regions and are easily accessible, the CMIP5 ensemble is more convenient for regional hydrological assessments than the perturbed physics experiments.

The Sensitivity of Heavy Precipitation to Horizontal Resolution, Domain Size, and Rain Rate Assimilation: Case Studies with a Convection-Permitting Model

The Australian Community Climate and Earth-System Simulator (ACCESS) is used to test the sensitivity of heavy precipitation to various model configurations: horizontal resolution, domain size, rain rate assimilation, perturbed physics, and initial condition uncertainties, through a series of convection-permitting simulations of three heavy precipitation (greater than 200 mm day−1) cases in different synoptic backgrounds. The larger disparity of intensity histograms and rainfall fluctuation caused by different model configurations from their mean and/or control run indicates that heavier precipitation forecasts have larger uncertainty. A cross-verification exercise is used to quantify the impacts of different model parameters on heavy precipitation. The dispersion of skill scores with control run used as “truth” shows thatthe impacts of the model resolution and domain size on the quantitative precipitation forecast are not less than those of perturbed physics and initial field uncertainties in these not intentionally selected heavy precipitation cases. The result indicates that model resolution and domain size should be considered as part of probabilistic precipitation forecasts and ensemble prediction system design besides the model initial field uncertainty.

Projecting the release of carbon from permafrost soils using a perturbed physics ensemble

The soils of the Northern Hemisphere permafrost region are estimated to contain 1100 to 1500 Pg of carbon (Pg C). A substantial fraction of this carbon has been frozen and therefore protected from microbial decay for millennia. As anthropogenic climate warming progresses much of this permafrost is expected to thaw. Here we conduct perturbed physics experiments on a climate model of intermediate complexity, with an improved permafrost carbon module, to estimate with formal uncertainty bounds the release of carbon from permafrost soils by year 2100 and 2300. We estimate that by 2100 the permafrost region may release between 56 (13 to 118) Pg C under Representative Concentration Pathway (RCP) 2.6 and 102 (27 to 199) Pg C under RCP 8.5, with substantially more to be released under each scenario by year 2300. A subset of 25 model variants were projected 8000 years into the future under continued RCP 4.5 and 8.5 forcing. Under the high forcing scenario the permafrost carbon pool decays away over several thousand years. Under the moderate scenario forcing a remnant near-surface permafrost region persists in the high Arctic which develops a large permafrost carbon pool, leading to global recovery of the pool beginning in mid third millennium of the common era (CE). Overall our simulations suggest that the permafrost carbon cycle feedback to climate change will make a significant but not cataclysmic contribution to climate change over the next centuries and millennia.