Bushfire fuel hazard is determined by fuel hazard that represents the type, amount, density, and three-dimensional distribution of plant biomass and litter. The fuel hazard represents a biological control on fire danger and may change in future with plant growth patterns. Rising atmospheric CO2 concentration (Ca) tends to increase plant productivity (‘fertilisation effect’) but also alters climate, leading to a ‘climatic effect’. Both effects will impact on future vegetation and thus fuel hazard. Quantifying these effects is an important component of predicting future fire regimes and evaluating fire management options. Here, by combining a machine learning algorithm that incorporates the power of large fine-resolution datasets with a novel optimality model that accounts for the climatic and fertilisation effects on vegetation cover, we developed a random forest model to predict fuel hazard at fine spatial resolution across the state of Victoria in Australia. We fitted and evaluated model performance with long-term (i.e., 20 years), ground-based fuel observations. The model achieved strong agreement with observations across the fuel hazard range (accuracy >65%). We found fuel hazard increased more in dry environments to future climate and Ca. The contribution of the ‘fertilisation effect’ to future fuel hazard varied spatially by up to 12%. The predictions of future fuel hazard are directly useful to inform fire mitigation policies and as a reference for climate model projections to account for fire impacts.

Grassland responses to drought are strongly mediated by leaf phenology, with greening and browning being highly sensitive to soil moisture. However, this process is represented overly simplistically in most vegetation models, limiting their capacity to predict grassland responses to global change factors. We derive functions representing grassland phenological responses to soil water content (SWC), by fitting an empirical model to greenness data. Data were obtained from fixed cameras (phenocams) monitoring phenology at several grassland experiments in Sydney, Australia. The data-model synthesis showed that the sensitivity of growth to SWC exhibited a concave-down response in most species. For senescence, we found a strong nonlinear increase in senescence rate with declining SWC. Both findings contradict common assumptions in vegetation models. Incorporating nonlinear responses in the empirical model reduced the error in cover predictions by 12%. Model evaluation against data from drought treatments indicated that differential sensitivity of phenology to SWC helps explain differences among species’ responses to variable rainfall. Our work provides a new methodology, and new evidence, to support the development of improved representations of grassland phenology for vegetation models.