Please see attached pdf.

REVIEW OF MANUSCRIPT NHESS-2023-211

“On the potential of using smartphone sensors for wildfire hazard estimation”

Summary and recommendation

The manuscript presents a study of using smartphone-based T/RH measurements to estimate VPD and subsequently assess wildfire hazard. The use of smartphone-based data is certainly a novel aspect of the work presented and something that the wildfire community could examine in the future for assessing wildfire hazard or/and fire danger. However, despite the arguments presented by the authors in the manuscript, I, unfortunately, remain skeptical towards the added value of the proposed approach. I recommend to reconsider the manuscript for publication subject to a major revision that, in my view, would allow for better highlighting any possible added value of the proposed use of smartphone-based data for assessing the conduciveness of weather to wildfires.   

Major remarks

1. In L41, the authors claim that smartphone microsensor data “ay provide additional and highly complimentary data”, as compared to traditional weather station data. Further, in L91-93, the authors state that “smartphone sensors have the potential for providing useful information about VPD at high spatial resolution and high temporal resolution even in remote areas with few official weather stations”.

I am very worried that the analysis conducted by the authors and the results presented do not support the above two arguments made to support the added value of using smartphone data.

First, the authors conduct their analysis on a 1 x 1 degree grid, on which they interpolate the smartphone-based VPD estimates. This spatial resolution is too coarse for any assessment of wildfire hazard/danger and hence, the presented results cannot be used for supporting the authors’ claim about smartphone microsensors providing information about VPD at high spatial resolution.

Second, the authors computed gridded VPD data by taking the daily mean values of T and RH. This is a rather crude approach that introduces significant implications. By considering the daily mean T/RH values, the authors put together data that may have been collected both outdoors (daytime) and indoors (nighttime). In addition, there have been recent studies highlighting the importance of nighttime VPD for the effectiveness of fire suppression operations. During nighttime, fuels may recover part of their moisture content; when VPD remains high at night, this process is hindered and fuels’ flammability remains high, thus supporting the rapid spread of fire. By taking a daily mean T/RH value to estimate a kind of daily mean VPD, this important information is neglected.

Third, smartphone data availability is directly related to the population density. Where more people gather, more data will be available. This typically includes populated areas and very less often remote areas. In addition, when the smartphone data are gathered from people located in an urban area, one should take into account that conditions may be significantly different from a remote, mountainous area. By using these data to assess conditions in another totally different environment (in terms of land cover, topography, etc.), one should be aware of the significant uncertainty introduced through the extrapolation of the data.

In summary, I highly encourage the authors to reconsider the entire structure of their work to better highlighted the possible added value of smartphone data for wildfire applications. Some suggestions may include:

• Decrease the spatial resolution of the grid on to which the smartphone data are interpolated, possibly to match the resolution of one of the publicly available reanalysis datasets (e.g., ERA5 at 0.25 x 0.25 degrees).
• Comparison of the smartphone-based VPD against ERA5-based (or any other data source) VPD and VPD obtained from interpolation from weather stations. This would allow evaluating whether smartphones can provide additional information or not.
• Refrain from averaging daily T/RH values and focus on examining nighttime/daytime data.