A Bayesian stochastic generator to complement existing climate change scenarios: supporting uncertainty quantification in marine and coastal ecosystems

Abstract

Available climate change projections, which can be used for quantifying future changes in marine and coastal ecosystems, usually consist of a few scenarios. Studies addressing ecological impacts of climate change often make use of a low- (RCP2.6), moderate- (RCP4.5) or high climate scenario (RCP8.5), without taking into account further uncertainties in these scenarios. In this research a methodology is proposed to generate further synthetic scenarios, based on existing datasets, for a better representation of climate change induced uncertainties. The methodology builds on Regional Climate Model scenarios provided by the EURO-CORDEX experiment. In order to generate new realizations of climate variables, such as radiation or temperature, a hierarchical Bayesian model is developed. In addition, a parameterized time series model is introduced, which includes a linear trend component, a seasonal shape with varying amplitude and time shift, and an additive residual term. The seasonal shape is derived with the non-parametric locally weighted scatterplot smoothing, and the residual term includes the smoothed variance of residuals and independent and identically distributed noise. The distributions of the time series model parameters are estimated through Bayesian parameter inference with Markov chain Monte Carlo sampling (Gibbs sampler). By sampling from the predictive distribution numerous new statistically representative synthetic scenarios can be generated including uncertainty estimates. As a demonstration case, utilizing these generated synthetic scenarios and a physically based ecological model (Delft3D-WAQ) that relates climate variables to ecosystem variables, a probabilistic simulation is conducted to further propagate the climate change induced uncertainties to marine and coastal ecosystem indicators.

Artificial Intelligence for Earth System Predictability (AI4ESP) (2021 Workshop Report)

“…Generative models, a different class of stochastic emulator, seek to quantify aleatoric uncertainty (such as internal/chaotic natural variability) and sample random realizations of this uncertainty. These may include statistical models (Vesely et al 2019;Link et al 2019;Mészáros et al 2021;Verdin et al 2019), dynamical reduced models (Foster, Comeau, and Urban 2020), variational autoencoders (Tibau et al 2021), and normalizing flows (Groenke, Madaus, and Monteleoni 2020). Dunbar et al (2021), Berdahl et al (2021), andBeusch, Gudmundsson, andSeneviratne (2020) have all utilized a Gaussian process emulator approach for the calibration of an idealized global climate model (GCM) and for the CISM ice sheet model, respectively (see Figure 12-1), Cleary et al ( 2021) also proposed a calibrate-emulate-sample approach using GPs, and Watson-Parris et al ( 2021) have released open-source software for Earth system emulation, which is built on top of GPyTorch.…”

Climate Change Induced Trends and Uncertainties in Phytoplankton Spring Bloom Dynamics

“…Since the RCM simulations are subject to climate model structural error and boundary errors from the driving GCMs (Navarro-Racines et al, 2020), they should be bias corrected before applying them in impact studies (Luo, 2016). For this reason, quantile mapping bias correction (Amengual et al, 2012) was applied using the RCM simulations for the reference period and daily historical field measurements from KNMI for the same period, as described in Mészáros et al (2021). The quantile-quantile mapping transfer functions were established for the reference period and separately for each RCM simulation.…”

“…Consequently, to better characterize uncertainties, an enriched set of climate change projections is employed. This set of air temperature and solar radiation projections was produced using a Bayesian stochastic generator (Mészáros et al, 2021), which builds on the above mentioned Regional Climate Model scenarios provided by the EURO-CORDEX experiment and generates further synthetic scenarios using a hierarchical Bayesian model. The generated ensemble of air temperature and solar radiation projections include 120 members and their statistical properties are similar to the input projections.…”

“…This leads us to the second reason which might explain the lack of difference in cardinal dates between RCPs. The generated scenarios have been produced with a Bayesian stochastic generator introduced in Mészáros et al (2021). This model assumes that Euro-CORDEX scenarios are exchangeable rather FIGURE 14 | Projected future phytoplankton spring bloom beginning (A), peak timing (B), and end (C) under generated (G) radiation and temperature projections (based EURO-CORDEX) .…”