Abstract:With the underlying aim of increasing efficiency of computational modelling pertinent for managing & protecting the Great Barrier Reef, we perform a preliminary investigation on the use of deep neural networks for opportunistic model emulation of APSIM models by repurposing an existing large dataset containing outputs of APSIM model runs. The dataset has not been specifically tailored for the model emulation task. We employ two neural network architectures for the emulation task: densely connected feed-forward neural network (FFNN), and gated recurrent unit feeding into FFNN (GRU-FFNN), a type of a recurrent neural network. Various configurations of the architectures are trialled. A minimum correlation statistic is used to identify clusters of APSIM scenarios that can be aggregated to form training sets for model emulation. We focus on emulating 4 important outputs of the APSIM model: runoff, soil_loss, DINrunoff, Nleached. The GRU-FFNN architecture with three hidden layers and 128 units per layer provides good emulation of runoff and DINrunoff. However, soil_loss and Nleached were emulated relatively poorly under a wide range of the considered architectures; the emulators failed to capture variability at higher values of these two outputs. While the opportunistic data available from past modelling activities provides a large and useful dataset for exploring APSIM emulation, it may not be sufficiently rich enough for successful deep learning of more complex model dynamics. Design of Computer Experiments may be required to generate more informative data to emulate all output variables of interest. We also suggest the use of synthetic meteorology settings to allow the model to be fed a wide range of inputs. These need not all be representative of normal conditions, but can provide a denser, more informative dataset from which complex relationships between input and outputs can be learned.

Don't Waste Data: Transfer Learning to Leverage All Data for Machine-Learnt Climate Model Emulation

Achieving Robust and Transferable Performance for Conservation‐Based Models of Dynamical Physical Systems

ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling

Transfer Learning for Emulating Ocean Climate Variability across $CO_2$ forcing

Machine learning emulation of a local-scale UK climate model

Predicting global patterns of long-term climate change from short-term simulations using machine learning

Efficient Climate Simulation via Machine Learning Method

High-resolution meteorology with climate change impacts from global climate model data using generative machine learning

Towards Learned Emulation of Interannual Water Isotopologue Variations in General Circulation Models

Can Transfer Learning be Used to Identify Tropical State-Dependent Bias Relevant to Midlatitude Subseasonal Predictability?

A Deep Learning Earth System Model for Stable and Efficient Simulation of the Current Climate

Evaluating Loss Functions and Learning Data Pre-Processing for Climate Downscaling Deep Learning Models

Generalized Online Transfer Learning for Climate Control in Residential Buildings

Evaluating the transferability potential of deep learning models for climate downscaling

Machine Learning Climate Model Dynamics: Offline versus Online Performance

Training machine learning models on climate model output yields skillful interpretable seasonal precipitation forecasts

A probabilistic framework for learning non-intrusive corrections to long-time climate simulations from short-time training data

Data-Driven Weather Forecasting and Climate Modeling from the Perspective of Development

Data Imbalance, Uncertainty Quantification, and Generalization via Transfer Learning in Data-driven Parameterizations: Lessons from the Emulation of Gravity Wave Momentum Transport in WACCM

Probabilistic Emulation of a Global Climate Model with Spherical DYffusion

Opportunistic Emulation of Computationally Expensive Simulations via Deep Learning