Post-fire mortality models and satellite-derived severity indices are the primary methods available to assess fire effects across broad scales, and these tools are foundational to the monitoring and management of fire-prone forest ecosystems throughout the west. Despite the widespread use of these tools, however, there is a persistent need to improve their accuracy, determine their applicability in different forest types, and ensure they remain relevant under uncertain climate futures.

I evaluated the performance of mortality models within the First Order Fire Effects Model (FOFEM) software, and compared their performance to locally-parameterized models based on five different forms. I assessed model accuracy at the individual-tree, stand, and community levels, and compared stand-level accuracy across a range of spatial scales. FOFEM consistently under-predicted mortality for the three conifer species, possibly because of the timing of the fire during a severe multi-year drought. Model error was spatially correlated at small spatial scales (<2.5 ha), suggesting that mortality was mediated by spatially-structured delayed mortality processes.

I also conducted a comparison of 36 satellite-derived spectral indices using tree-based measures of fire severity and estimated uncertainty in actual fire effects. dNBRperformed reliably well, but accuracy could be improved by using other indices to detect specific aspects of tree mortality (percent basal area mortality versus percent stem mortality). Relativized versions of dNBRdid not consistently improve accuracy; the relativized burn ratio (RBR) was generally equivalent to dNBR, but RdNBRhad consistently lower accuracy. There was a considerable amount of uncertainty at the landscape scale, with an expected range in estimated percent basal area mortality greater than 37% for half of the area burned (>50,000 ha).

This research employed both familiar and novel analytical techniques to empirically evaluate these essential fire science tools and identify ways in which they may be advanced. I used a large, intensively sampled forest demography plot to assess error across multiple spatial scales, among different tree diameter classes, and with a variety of performance metrics. The unique nature of this dataset provided a new perspective through which these tools could be evaluated, and this revealed unexpected results that contribute to both theoretical and applied aspects of fire science.

This presentation was part of the 2018-2019 Firelab Seminar Series. Video recording is available from the Fifirelab web site at www.firelab.org