An official website of the United States government
Abstract Parameters in climate models are usually calibrated manually, exploiting only small subsets of the available data. This precludes both optimal calibration and quantification of uncertainties. Traditional Bayesian calibration methods that allow uncertainty quantification are too expensive for climate models; they are also not robust in the presence of internal climate variability. For example, Markov chain Monte Carlo (MCMC) methods typically requiremodel runs and are sensitive to internal variability noise, rendering them infeasible for climate models. Here we demonstrate an approach to model calibration and uncertainty quantification that requires onlymodel runs and can accommodate internal climate variability. The approach consists of three stages: (a) a calibration stage uses variants of ensemble Kalman inversion to calibrate a model by minimizing mismatches between model and data statistics; (b) an emulation stage emulates the parameter‐to‐data map with Gaussian processes (GP), using the model runs in the calibration stage for training; (c) a sampling stage approximates the Bayesian posterior distributions by sampling the GP emulator with MCMC. We demonstrate the feasibility and computational efficiency of this calibrate‐emulate‐sample (CES) approach in a perfect‐model setting. Using an idealized general circulation model, we estimate parameters in a simple convection scheme from synthetic data generated with the model. The CES approach generates probability distributions of the parameters that are good approximations of the Bayesian posteriors, at a fraction of the computational cost usually required to obtain them. Sampling from this approximate posterior allows the generation of climate predictions with quantified parametric uncertainties.
more »
« less
Ensemble‐Based Experimental Design for Targeting Data Acquisition to Inform Climate Models
Abstract Data required to calibrate uncertain general circulation model (GCM) parameterizations are often only available in limited regions or time periods, for example, observational data from field campaigns, or data generated in local high‐resolution simulations. This raises the question of where and when to acquire additional data to be maximally informative about parameterizations in a GCM. Here we construct a new ensemble‐based parallel algorithm to automatically target data acquisition to regions and times that maximize the uncertainty reduction, or information gain, about GCM parameters. The algorithm uses a Bayesian framework that exploits a quantified distribution of GCM parameters as a measure of uncertainty. This distribution is informed by time‐averaged climate statistics restricted to local regions and times. The algorithm is embedded in the recently developed calibrate‐emulate‐sample framework, which performs efficient model calibration and uncertainty quantification with onlymodel evaluations, compared withevaluations typically needed for traditional approaches to Bayesian calibration. We demonstrate the algorithm with an idealized GCM, with which we generate surrogates of local data. In this perfect‐model setting, we calibrate parameters and quantify uncertainties in a quasi‐equilibrium convection scheme in the GCM. We consider targeted data that are (a) localized in space for statistically stationary simulations, and (b) localized in space and time for seasonally varying simulations. In these proof‐of‐concept applications, the calculated information gain reflects the reduction in parametric uncertainty obtained from Bayesian inference when harnessing a targeted sample of data. The largest information gain typically, but not always, results from regions near the intertropical convergence zone.
more »
« less
- Award ID(s):
- 1835860
- PAR ID:
- 10373782
- Publisher / Repository:
- DOI PREFIX: 10.1029
- Date Published:
- Journal Name:
- Journal of Advances in Modeling Earth Systems
- Volume:
- 14
- Issue:
- 9
- ISSN:
- 1942-2466
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract In this paper, we are interested in the following question: given an arbitrary Steiner triple systemonvertices and any 3‐uniform hypertreeonvertices, is it necessary thatcontainsas a subgraph provided? We show the answer is positive for a class of hypertrees and conjecture that the answer is always positive.more » « less
-
Abstract Global ocean mean salinityis a key indicator of the Earth's hydrological cycle and the exchanges of freshwater between land and ocean, but its determination remains a challenge. Aside from traditional methods based on gridded salinity fields derived from in situ measurements, we explore estimates ofbased on liquid freshwater changes derived from space gravimetry data corrected for sea ice effects. For the 2005–2019 period analyzed, the differentseries show little consistency in seasonal, interannual, and long‐term variability. In situ estimates show sensitivity to choice of product and unrealistic variations. A suspiciously large rise insince ∼2015 is enough to measurably affect halosteric sea level estimates and can explain recent discrepancies in the global mean sea level budget. Gravimetry‐basedestimates are more realistic, inherently consistent with estimated freshwater contributions to global mean sea level, and provide a way to calibrate the in situ estimates.more » « less
-
Abstract Bin and bulk schemes are the two primary methods to parameterize cloud microphysical processes. This study attempts to reveal how their structural differences (size‐resolved vs. moment‐resolved) manifest in terms of cloud and precipitation properties. We use a bulk scheme, the Arbitrary Moment Predictor (AMP), which uses process parameterizations identical to those in a bin scheme but predicts only moments of the size distribution like a bulk scheme. As such, differences between simulations using AMP's bin scheme and simulations using AMP itself must come from their structural differences. In one‐dimensional kinematic simulations, the overall difference between AMP (bulk) and bin schemes is found to be small. Full‐microphysics AMP and bin simulations have similar mean liquid water path (mean percent difference <4%), but AMP simulates significantly lower mean precipitation rate (−35%) than the bin scheme due to slower precipitation onset. Individual processes are also tested. Condensation is represented almost perfectly with AMP, and only small AMP‐bin differences emerge due to nucleation, evaporation, and sedimentation. Collision‐coalescence is the single biggest reason for AMP‐bin divergence. Closer inspection shows that this divergence is primarily a result of autoconversion and not of accretion. In full microphysics simulations, lowering the diameter threshold separating cloud and rain category in AMP fromtoreduces the largest AMP‐bin difference to ∼10%, making the effect of structural differences between AMP (and perhaps triple‐moment bulk schemes generally) and bin even smaller than the parameterization differences between the two bin schemes.more » « less
-
Abstract We study the stability and nonlinear local dynamics of spectrally stable periodic wave trains of the Korteweg‐de Vries/Kuramoto‐Sivashinsky equation when subjected to classes of periodic perturbations. It is known that for each, such a‐periodic wave train is asymptotically stable to‐periodic, i.e., subharmonic, perturbations, in the sense that initially nearby data will converge asymptotically to a small Galilean boost of the underlying wave, with exponential rates of decay. However, both the allowable size of initial perturbations and the exponential rates of decay depend onand, in fact, tend to zero as, leading to a lack of uniformity in such subharmonic stability results. Our goal here is to build upon a recent methodology introduced by the authors in the reaction–diffusion setting and achieve a subharmonic stability result, which is uniform in. This work is motivated by the dynamics of such wave trains when subjected to perturbations that are localized (i.e., integrable on the line).more » « less
