Managing fire, especially wildfire, in grassland suffers currently from the inability to anticipate large wildfires, especially fuel curing, when live herbaceous vegetation transitions rapidly to dead fuel. To assess these herbaceous fuel dynamics in grassland, we conducted three studies: 1) a study that used a database of large wildfires in Oklahoma to examine the relationship of fire occurrence and fire size with soil moisture, 2) an intensive field-based study to quantify and subsequently model herbaceous fuel load and moisture content in grassland patches that differed in time since fire and, therefore, proportion of live and dead herbaceous fuel load, and 3) modeling the influence of herbaceous fuel dynamics and weather conditions on fire behavior in tallgrass prairie. Wildfire occurrence in Oklahoma exhibits seasonally-dependent relationships with soil moisture conditions. Large growing-season wildfires in Oklahoma occur almost exclusively under conditions of low soil moisture, and growing-season fire probability is particularly high when less than 20% of the soil’s available water capacity is filled. Dormant-season wildfire occurrence is also related to soil moisture conditions, although not as strongly. Both current soil moisture levels and soil moisture levels during the previous growing season influence wildfire probability during the growing season. The physical link between soil moisture and vegetation moisture, along with the increasing availability of soil moisture data, make soil moisture a strong candidate variable for monitoring wildfire risk, especially in the growing season. In the intensive field study, we collected fuelbed data from a grassland research location with successional vegetation produced by fuel treatments of spatially and temporally variable fire and grazing. The data collected (and variables estimated from the data) include a suite of fuel variables (live and dead load, live and dead moisture content, particle density, etc.), soil moisture, soil temperature, and plant canopy spectral reflectance from a hand-held radiometer and from satellite. Results from our study on the relationship of soil moisture and wildfire occurrence informed our approach to modeling fuel dynamics. Candidate variables will focus on soil moisture and canopy reflectance to model fuel components that our research shows vary most over space and time. Modeling fire behavior across the full range of potential fire conditions, including fuel characteristics, revealed complex behaviors that have not been accounted for in grassland fire management. We found that fire behavior changes in complex ways as grass curing progresses. Linear relationships became non-linear, and tipping points occurred that revealed rapid and sudden change in fire behavior at intermediate levels of grass curing.