Vertical windspeed was enhanced over ambient conditions during each fire, with maximum 10 hz values reaching 3.7, 3.8, and 8.3 m s-1 four meters above the canopy during each prescribed fire. Maximum air temperature four meters above the canopy estimated from 10 hz sonic anemometer speed of sound measurements was 32.3, 44.2, and 117.8 °C. 10 hz vertical windspeed and air temperature were positively correlated as the fire front passed under the tower at each fire, with correlations highest at the hottest burn (r2 = 0.48). Total energy flux, calculated as the sum of latent and sensible heat fluxes above the canopy for the two cooler burns, was nine and five times greater than available energy estimated from net radiation and soil heat flux measurements; Rnet G. Half-hourly sensible heat flux peaked at 1762 and 1675 W m-2, and water vapor flux at 524 and 483 W m-2, respectively. Total energy release during these fires calculated from the excess H and LE flux after correction for available energy was 6,389 and 7,256 kJ m-2.
Fuel consumption on the forest floor and in the understory totaled 8.2, 9.8 and 7.1 metric tons ha-1, representing 46, 44 and 49 % of pre-burn loadings, respectively. Corresponding heat of combustion values, calculated assuming complete combustion of consumed fuels at measured water contents, were 12,991, 15,422, and 11,280 kJ m-2 for each prescribed burn, respectively. Flux measurements totaled a smaller fraction of energy release during combustion, because energy was consumed in the heating of the surrounding air and soil, and radiant energy is likely under sampled using eddy covariance sensors. It also is possible that the flux measurements only sampled a limited portion of the plume, because we may not have burned as intensely directly under the tower compared to other areas in the burn block. In addition, 10 hz data may underestimate instantaneous fluxes during enhanced turbulent transfer occurring in fires, because smoke occasionally interfered with the sonic sensors, potentially dampening higher speed turbulence signals, and the LiCor LI-7000 used to sample water vapor may not accurately sample such large fluctuations in H20 concentrations. Quantifying consumption using pre- and post-burn field plots also is not without error, and the calculated SD for consumption of 1-hour fuels represents 23-32% of the total. In addition, char particles < 2 mm diameter that were produced from litter during the prescribed fire were not sampled, because we sifted samples through 2 mm mesh size screens to remove sand and fine grained organic matter.
Despite sampling limitations, simultaneous quantification of fluxes and fuel consumption during fires are essential for evaluating predictive plume dispersion models such as the Weather Research and Forecasting (WRF) FLEXPART system, the Regional Atmospheric Modeling System Forest Large Eddy Simulation (RAFLES) system, and fuel combustion models such as CONSUME3.