fldgen

fldgen generates synthetic climate variable fields for temperature and precipitation by learning and reproducing spatial, temporal, and inter-variable correlation structures from Earth System Models (ESMs) to enable large ensembles for climate impact analyses.

Key Features:

• Joint Realization of Multiple Variables: Produces joint realizations of multiple climate variables, primarily temperature and precipitation, for multivariate analyses.
• Internal Variability and Correlation Modeling: Learns and replicates spatial, temporal, and inter-variable correlation structures from ESM outputs to generate random two-variable fields that preserve variance and covariance.
• Efficiency and Computational Economy: Operates at low computational cost to enable rapid generation of thousands of climate field realizations suitable for analyzing low-frequency, high-impact events such as multi-year droughts.
• Removal of Normal Distribution Requirement: Generates covarying temperature and precipitation data without requiring variables to be normally distributed.

Scientific Applications:

• Climate Change Impact Studies: Generates extensive synthetic datasets to assess potential climate change impacts on human and natural systems.
• Scenario Analysis: Provides synthetic but realistic data to support analysis of different climate pathways and scenario ensembles.
• Multi-variable Climate Modeling: Enables study of interactions between temperature and precipitation through joint realizations, supporting multivariate predictive modeling.

Methodology:

Learns from an ESM's variability patterns and generates synthetic data that retain the original model's spatial and temporal correlation structures; this process involves randomizing residuals of pattern-scaling temperature outputs (or other climate variables) to produce realistic field realizations.