This manuscript presents an interesting evaluation of two methods, one based on "fire effective temperature" and another using a potassium emission line (K-line), for retrieving spatially explicit estimates of the combustion phase of combusting biomass. Using the latter approach, which was generally superior, three models are developed to assess the degree to which the retrieved combustion phase is able to improve the accuracy of smoke emission rates, smoke composition, and the overall rate of fuel consumption.

References are adequate and appropriate. The topic is within the scope of AMT. The work appears sound and represents a significant and useful contribution to fire emissions estimation and our understanding of those emissions. I recommend publication after several minor clarifications and corrections have been made.

Line 95: Suggest changing "...the smoke production process:" to "...the smoke production process, defined as:"

Lines 105-108: A clarification of what "effective temperature" (T_f) actually means is warranted here since emissivity does not appear in Eqs. (2) and (3). This implies that T_f is a radiant temperatures rather than actual fire temperature as measured with a thermometer, but the intended meaning is not clear from the text.

Eq. 2 and 3: Please make mathematical notation here and in text consistent, e.g. within Eq. 2 and 3 the variable p is sometimes italicized and sometimes not.

Eq. 2 and 3: As the effective temperature temperature is define, 

Eqs. 4 and 17: Here also please use italics consistently for variables and constants - see comment above in reference to Eqs. 2 and 3.

Line 147: Suggest using Oxford comma for clarity: "oak kindling, pine forest litter, and soybean crop residue."

Line 163: Symbols/characters between "each image that had" and "600 K" seem to be garbled.

Line 171: Suggest writing "(L_FD+L_SD+L_C)" as "(L_FD, L_SD, and L_C)" for clarity.

Line 369: Not clear what the uncertainty (0.28) attached to the mean m_k represents or how it was calculated.

Eqs. (15): I think the condition here should be >= to be consistent with the K-line detection described in section 4.1.

Line 385: Here and later in text change "timeseries" to "time series".

Fig. 7: Please state approximate spatial dimensions of panel (b).

Fig. 8: Based on the panels above I would expect the MCE to show a bit more spatial variability. Would a nonlinear color scale possibly reveal more features?

Supplement Figs. S1-S11: Change "Times series" to "Time series" in captions.

The manuscript describes a series of laboratory fuel combustion experiments which investigate the capacity of remotely sensed data to detect flaming and smouldering phases of combustion. The experiments illustrate the variation of the rate trace gas emissions and FRP with combustion phase and highlight the capacity of K-line spectral measurements to identify flaming combustion.

Combustion phase specific emissions models are developed from these experiments and applied to airborne hyperspectral data over a Canadian wildfire resulting in improved emissions estimates when utilising FRP and K-line data over those using a ‘fire averaged’ emissions coefficient applied to FRP retrievals.

The research is interesting, novel and adds to the body of work on wildfire emissions estimation. The manuscript will be of interest to readers of the journal. The manuscript is suitable for publication subject to a few minor edits outlined below.

Comments

#line 21 – define FRP acronym

#line 71 – ‘(Zhang et al., 2015)’

#line 163 – typo - ‘T  ³ 600 K’

#line 213 – ‘measure’

#Fig 2 – Closing bracket missing on plot y-axis

#line 322 – define FREM acronym

#Figure 5/line 301– for clarity it would be beneficial to include the percentage of observations which were detected as containing flaming combustion (e.g. AKBD > 1.5). It appears to be the majority in the plots although in reality most observations are in the smouldering phase (e.g. Fig 3) which has less variation in values.

#line 430 what is the spatial resolution of the hyperspectral data used in this analysis? To what extent does spatial resolution influence the detection of pixels containing flaming activity?

References – a couple of references are missing details (e.g.  Magidimisha et al and  Urbanski, S.)

Supplementary material

In some plots the legend would benefit from being repositioned or reducing the text in the legend to avoid overlapping the data