Starflock

https://github.com/GaelleMG/starflock_firexaq

https://docs.google.com/presentation/d/1gt_6hwCpTVRXQCnGRerzhzYUWaX8EAeqvhPFEo4vGqI/edit?usp=sharing

Background

For this challenge, we were interested to learn about the difference in the composition of smoke that originated from wildfires as compared to agricultural burns. We suspected that due to agricultural practices such as fertilization and addition of other chemicals the aerosol composition of smoke would differ. With this knowledge, we could then create a website or app that helps the public understand the risk factors of each type of fire as well as help farmers make more informed agricultural management decisions.

Wildfires and agricultural burning cost the economy billions. What if this could be avoided by better predicting when they were going to happen and detecting the signs earlier? Now, more than ever is the time to solve for this and better inform the public about air quality by pulling in historical and informative data available from NASA, while also leveraging our understanding of aerosol/ fire physiology to create a safer and healthier society.

Phase I

Understand the FIREX-AQ DC8 mission and corresponding data; access and explore trends

The mission of FIREX-AQ (Fire Influence on Regional to Global Environments and Air Quality) according to NASA/NOAA is: "The overarching objective of FIREX-AQ is to provide measurements of trace gas and aerosol emissions for wildfires and prescribed fires in great detail, relate them to fuel and fire conditions at the point of emission, characterize the conditions relating to plume rise, follow plumes downwind to understand chemical transformation and air quality impacts, and assess the efficacy of satellite detections for estimating the emissions from sampled fires."

There are many datasets that are publicly available, so our first task was to select several to explore. One dataset was selected from the DC8 instrument plane missions. Data from all flights (23 total) were downloaded from the NASA website (see link below). A total of approximately 20 million rows of data were used for our analyses.

First we normalized all aerosol measurements to the density of organic aerosols at the time of sampling. This prevents changes in measurements from being inappropriately attributed to a difference in composition when in fact it may have been due to the amount of smoke in the sample taken during flight.

We then plotted the normalized aerosol measurements by altitude. Below are representative plots for three flights collecting wildfire smoke followed by three flights collected agricultural burn smoke.

WILDFIRE (Three large panels):

Agricultural Burn Fires (Three large panels):

We then plotted the normalized measurements by latitude and longitude. Below are some examples:

We also plotted the flight tracks to distinguish the WILDFIRE missions from the AGRI BURN missions as well as all the missions by mission ID:

Phase II

Analyze statistically the effect of each aerosol measure on predicting whether smoke originated from a wildfire or agricultural burn.

Using logistical regression, we tested the contribution of each normalized measurement on the probability that the smoke originated from a wildfire or agricultural burn by labeling each row of measurement based on each flight mission.

According to the above analysis, many aerosols were significantly different and able to help predict smoke origin. Of interest, ammonium and nitrate were predictors. This would reflect the use of fertilizers in agriculture. Below is a summary of each aerosol in the above list plotted by fire origin type (agriburn vs. wildfire):

Phase III

Create a platform to share the findings that is accessible to the public.

With more time we will optimize our App with real-time data pulled from various satellites, agricultural departments and various API's (crowdsourcing) available on air quality. This will allow for enhanced historical referencing, and increased the accuracy of the final fire prediction of our App. For more details on the app, read below:

The goal of our version 1 app is to (1) help citizens understand in a 'real-time' manner emergency information on the fires in their area and (2) detect differences between various fire-types based on the aerosol physiology. It will pull in coordinates via Google Maps, and leverage the smartphone camera to detect fire color wavelength.

With the data input from the everyday citizen, we can then leverage the historical datasets collected (and our own interpretation of their statistical significance on where typical agriculture fires have occurred vs. wildfires, based on the chemical data taken from NASA) to: (1) educate the citizen on the nature of the fire (percentage that the fire is x miles away based on historical's, and if it is either an agriculture (chemical) fire or wildfire) and (2) leverage the aerosol optical depth information from NASA to help (a) predict the seriousness of the fire and (b) further provide accuracy for it's fire-type. For example, a chemical fire will likely have green or orange tones, based on the interaction with agricultural fertilizers vs. there would be more white smoke in a wildfire as it could be nearby water. The app will then predict the type of fire nearby with a percentage of accuracy. What's great is we can help educate the public on what to look for as well in agriculture burns vs. wildfires.

Our version 2.0 app will eventually pull in information from agricultural data. With time we can increase the strength of our App offering and help reduce crop waste by overlaying aerosol and fire physiology data with crop yield. The potential is endless and will help society stay safe but also save the economy potentially billions in crop damages from smoke and fires.

NASA FIREX-AQ Airbone Data (From the Jimenez team):

https://www-air.larc.nasa.gov/cgi-bin/ArcView/firexaq

Interesting! We all have suffered in one way or another from the impacts of global warming and resulting wildfires. Especially today with the cost of food going up due to inflation, society doesn't need another reason for supply chain to be impacted and costs going up further. We decided to put our thinking hats on with our unique backgrounds in Chemistry and Data/Science to better understand the impact the chemistry of fire has on how fires develop and spread, with the goal to reduce harvest waste and better predict their spread.

The hardest part was sifting through all the data and making sense of it, but we got there.

See Github link above for references

#airborne #firexaq #dataviz #smoke #wildfire #agriculture