A semi-mechanistic model for predicting the moisture content of fine litter
The moisture content of vegetation and litter (fuel moisture) is an important determinant of fire risk, and predictions of dead fine fuel moisture content (fuel with a diameter <25.4 mm) are particularly important. A variety of indices, as well as empirical and mechanistic models, have been proposed to predict fuel moisture, but these approaches have seldom been validated across temporally extensive datasets, or widely contrasting vegetation types. Here, we describe a semi-mechanistic model, based on the exponential decline of fuel moisture content with atmospheric vapor pressure deficit, that predicts daily minimum fuel moisture content. We calibrated the model at one site in New South Wales, Australia, and validated it at three contrasting ecosystem types in California, USA, where 10-h fuel moisture content was continuously measured every 30 min over a year. We found that existing drought indices did not accurately predict fuel moisture, and that empirical and equilibrium models provided biased estimates. The mean absolute error (MAE) of the fuel moisture content predicted by our model across sites and years was 3.7%, which was substantially lower than for other, commonly used models. Our model's MAE dropped to 2.9% when fuel moisture was below 20%, and to 1.8% when fuel moisture was below 10%. Our model's MAE was comparable to instrumental MAE (3.1-2.5%), indicating that further improvement may be limited by measurement error. The simplicity, accuracy and precision of our model makes it suitable for a range of applications, such as operational fire management and the prediction of fire risk in vegetation models, without the need for site-specific calibrations.