Abstract:

Absorption processes of dead larch twigs with three different diameters(0.5, 1.0 and 1.5 cm) in laboratory were modeled by a method proposed by Catchpole et al.(2001). Robustness of the model was tested by cross validation. Errors occurred in the model extrapolation were also analyzed. Results showed that moisture-related parameters of timelag and equilibrium moisture content had small variation in the cross validation, which indicated the results were quite robust. Variety existed in timelag and parameters related to equilibrium moisture content among twigs with the same diameter. Although both the errors in applying the models to data on which the models were based and those in applying the models to data on which the models were not based followed normal distribution, the latter was larger than the former. The latter errors occurred with lower probability in sections where the former errors occurred with higher probability and also occurred in sections where the latter errors did not occur. This paper also presented conditional probability and means of errors under different predicted fuel moisture content, which indicated that not only the means but also the probability of larger errors increased with increasing predicted moisture. This probability could allow users of the models to estimate errors on a particular predicted moisture value, which can reduce uncertainty in fire danger rating and fire behavior modeling which uses the predicted moisture. The results also suggested that models established using pooled data from at least 4 twigs could reduce the uncertainty in the model extrapolation in fuel moisture prediction. Since the variation between the twigs and conditions is much less than those occurred in the field, the results can be regarded as lower limits of errors occurred in model extrapolation, which, nevertheless, can provide useful clues to assess model performance when it is extrapolated though full evaluation of the performance is still needed by further research.

Key words: larch, fuel, moisture content, robustness, extrapolation, error