Abstract:Neural networks are increasingly used in environmental science applications. Furthermore, neural network models are trained by minimizing a loss function, and it is crucial to choose the loss function very carefully for environmental science applications, as it determines what exactly is being optimized. Standard loss functions do not cover all the needs of the environmental sciences, which makes it important for scientists to be able to develop their own custom loss functions so that they can implement many of the classic performance measures already developed in environmental science, including measures developed for spatial model verification. However, there are very few resources available that cover the basics of custom loss function development comprehensively, and to the best of our knowledge none that focus on the needs of environmental scientists. This document seeks to fill this gap by providing a guide on how to write custom loss functions targeted toward environmental science applications. Topics include the basics of writing custom loss functions, common pitfalls, functions to use in loss functions, examples such as fractions skill score as loss function, how to incorporate physical constraints, discrete and soft discretization, and concepts such as focal, robust, and adaptive loss. While examples are currently provided in this guide for Python with Keras and the TensorFlow backend, the basic concepts also apply to other environments, such as Python with PyTorch. Similarly, while the sample loss functions provided here are from meteorology, these are just examples of how to create custom loss functions. Other fields in the environmental sciences have very similar needs for custom loss functions, e.g., for evaluating spatial forecasts effectively, and the concepts discussed here can be applied there as well. All code samples are provided in a GitHub repository.

Loss functions for spatial wildfire applications

Conceptualizing a probabilistic risk and loss assessment framework for wildfires

Uncertainty Aware Wildfire Management

Predicting burn probability: Dimensionality reduction strategies enable accurate and computationally efficient metamodeling

CIRA Guide to Custom Loss Functions for Neural Networks in Environmental Sciences -- Version 1

Wildfire Risk Prediction: A Review

Online Planning of Power Flows for Power Systems Against Bushfires Using Spatial Context

Reinforcement Learning for Wildfire Mitigation in Simulated Disaster Environments

Modelling wildland fire burn severity in California using a spatial Super Learner approach

Assessing and reinitializing wildland fire simulations through satellite active fire data

A Neural Emulator for Uncertainty Estimation of Fire Propagation

Integrating Pixel- and Polygon-Based Approaches to Wildfire Risk Assessment: Application to a High-Value Watershed on the Pike and San Isabel National Forests, Colorado, USA

An interpretable wildfire spreading model for real-time predictions

Modeling wildland fire burn severity in California using a spatial Super Learner approach

Experimental and Numerical Study on Data-Driven Prediction for Wildfire Spread Incorporating Adaptive Observation Error Adjustment

Modelling the Daily Probability of Wildfire Occurrence in the Contiguous United States

A Parallel Novelty Search Metaheuristic Applied to a Wildfire Prediction System

Deep graphical regression for jointly moderate and extreme Australian wildfires

Discrete Wildfire Simulation Case Study

Wildfires Vegetation Recovery through Satellite Remote Sensing and Functional Data Analysis