A machine learning approach to emulation and biophysical parameter estimation with the Community Land Model, version 5

Abstract

Abstract. Land models are essential tools for understanding and predicting terrestrial processes and climate–carbon feedbacks in the Earth system, but uncertainties in their future projections are poorly understood. Improvements in physical process realism and the representation of human influence arguably make models more comparable to reality but also increase the degrees of freedom in model configuration, leading to increased parametric uncertainty in projections. In this work we design and implement a machine learning approach to globally calibrate a subset of the parameters of the Community Land Model, version 5 (CLM5) to observations of carbon and water fluxes. We focus on parameters controlling biophysical features such as surface energy balance, hydrology, and carbon uptake. We first use parameter sensitivity simulations and a combination of objective metrics including ranked global mean sensitivity to multiple output variables and non-overlapping spatial pattern responses between parameters to narrow the parameter space and determine a subset of important CLM5 biophysical parameters for further analysis. Using a perturbed parameter ensemble, we then train a series of artificial feed-forward neural networks to emulate CLM5 output given parameter values as input. We use annual mean globally aggregated spatial variability in carbon and water fluxes as our emulation and calibration targets. Validation and out-of-sample tests are used to assess the predictive skill of the networks, and we utilize permutation feature importance and partial dependence methods to better interpret the results. The trained networks are then used to estimate global optimal parameter values with greater computational efficiency than achieved by hand tuning efforts and increased spatial scale relative to previous studies optimizing at a single site. By developing this methodology, our framework can help quantify the contribution of parameter uncertainty to overall uncertainty in land model projections.

Sensitivity Analysis of Biome-BGCMuSo for Gross and Net Primary Productivity of Typical Forests in China

“…The parameters that most influenced GPP and NPP were those controlling respiration and the assimilation rate of photosynthesis, such as MRpern, FLNR, and C:Nleaf. This result is consistent with the results from White (2000), Raj (2014), andDagon (2020). MRpern represents the maintenance respiration in kg C/day per kg of tissue N, which is directly related to maintenance respiration and further affects GPP and NPP (Ryan 1991).…”

A Review of Machine Learning Applications in Land Surface Modeling

“…Physics guided machine learning models have also used DL-ANN as the ML technique [49]. DL-ANNs are also useful in constructing emulators of LSMs in to characterize the uncertainties and parameter optimization [55]. Dew point temperature is an important parameter to assess surface humidity conditions, critical for agricultural purposes.…”

“…In a recent study, Dagon et al [55] investigated the biogeophysical parameter space of CLM5 and determined the sensitivity of parameters to get insight into the role of parameter choices on the overall model uncertainty. They implemented an ML approach to globally calibrate six parameters of CLM5, selected by a sensitivity analysis, to the observations of water and carbon fluxes.…”

“…Parametric uncertainty can be investigated, and parameters can be better optimized following Dagon et al [55]. However, we need to consider longer data records (now possible with the availability of computing resources) and more output variables.…”

Assessing the potential for simplification in global climate model cloud microphysics

“…More complexity has also its downsides: More parameterised processes lead to more parametric uncertainty which in turn scientists investigate and try to reduce with large scientific effort (e.g. Rougier et al (2009); Lee et al (2011); Yan et al (2015); Williamson et al (2015); Dagon et al (2020)). In fact, Reddington et al (2017) argue that "aerosolclimate models are close to becoming an overdetermined system with many interacting sources of uncertainty but a limited range of observations to constrain them", refering to the complexity in the representation of aerosols and their interaction with clouds.…”