The EO Dashboard is derived from a suite of different satellites from the EO observation (EO) programs of the three collaborating space agencies: NASA, ESA, and JAXA. Your challenge is to devise a way to visually fuse information from diverse Earth observation datasets of different missions.
Forest fires emerge due to a complex set of events. However this is difficult to assess properly how human activities have an impact on forest fires. We want to provide a platform where Data used to predict forest fires and human activities and mobilities can be viewed together to understand forest fires.
We developed an app that permits users to view data at the county level in California USA from february 2020 to June 2021.
Wild forest fire is an important natural disaster that requires a complex set of data to be properly monitored and predicted. Most of the data used focus on Precipitation, Soil Moisture, Temperature, Humidity, Winds, Vegetation, Topography. However understanding the impact of human activities and mobilities on forest fire or the inverse is quite tedious.
Our app permit such understanding, We highlight burned area with 3 sets of factors:
The set "Natural factors" focus on data such as Soil Moisture, Precipitation, Temperature.
The set "Air factors" focus on data that can relate to human activities; nitrogendioxide (NO2), carbon monoxide (CO2), aerosol optical depth (AOD).
The set Mobility factors focus on human mobility in different places; grocery and pharmacy, parks, transit stations, workplaces, and residential.
To look at your app you can just go to this website. The map will zoom automatically on California. You can just click on the county you are interested in and look at the time series of the different factors. Please look at "demonstrate your solution" section for more information.
We hope that such app permit a better understanding of forest fire and human activities and mobilities. Such question can be answered using our app:
As an hillight we used the data to try to predict the burned area using a combination of different set of factors. By combining previous month data of the differents factors and the use of random forest we have seen a change in the features importance for the differents combination used.
In the table bellow, True indicate the usage of the data in the model as input data.
To sum up:
Nature -> Surface moisture anomaly, Surface temperature
Nature + Mobility -> transit_stations, residential
Nature + Air -> AOD, Surface moisture anomaly
Nature + AIR + Mobility -> AOD, transit_stations
Depending on the features used, the model gave different importance. Further study on the subject can be made using the app we created.
Recently one of us followed Nasa arset training on fire risk detection link. In some way covid19 permits us to study natural disasters without expecting too much human intervention at the precise time. How recent human activities have an impact in forest fire ? is forest fire only related to natural phenomena. Is forest fire decreasing with less human in parks ? All this unanswered questions were the main reason for our team to chose this challenge. A lot of data for multiple type of earth observations are required.
We decided to focus on 3 main factors. Natural factors that are the main data used to study forest fire:
Then Air data that relate to human activities as highlighted in no2 plumet in china or AOD_plumet_in_india:
Combined with Mobility data highlighted here. And finally the burned area as our main data to report forest fire.
We use Jupyter notebook and python for all the data collection and aggregation.
The app is a simple react app.It is bootstrapped with creat-react-app.The map is a mapbox map displayed using the library react-map-gl and deck.gl for the geojson of California.The charts are using the echarts library.It is deployed using netlify.
For the data collection part. The biggest problem was to have all the factors available for all the time stamps for each county. As we multiply the number of sources. Assuring that all the sources have correct data at the correct AOI is a problem. But we did it successfully using selected data sources and time ;(.
We mainly wanted to find interesting human data. For this we used the eodashboard at link. After some deliberation and seeing the differents economicals indicators available we decided to use the Mobility data in america.
Soil moisture anomaly : https://gimms.gsfc.nasa.gov/SMOS/SMAP/
Temperature : https://lpdaac.usgs.gov/products/mod11a1v006/
Precipitation : https://disc.gsfc.nasa.gov/datasets/GPM_3IMERGHH_06/summary
NO2 : https://sentinel.esa.int/web/sentinel/user-guides/sentinel-5p-tropomi
CO2 : https://sentinel.esa.int/web/sentinel/user-guides/sentinel-5p-tropomi
AOD : https://lpdaac.usgs.gov/products/mcd19a2v006/
Burned Area : https://lpdaac.usgs.gov/products/mcd64a1v006/
Mobility data : https://eodashboard.org/?indicator=GG&poi=GG-GG
None
Soil moisture anomaly : https://gimms.gsfc.nasa.gov/SMOS/SMAP/
Temperature : https://lpdaac.usgs.gov/products/mod11a1v006/
Precipitation : https://disc.gsfc.nasa.gov/datasets/GPM_3IMERGHH_06/summary
NO2 : https://sentinel.esa.int/web/sentinel/user-guides/sentinel-5p-tropomi
CO2 : https://sentinel.esa.int/web/sentinel/user-guides/sentinel-5p-tropomi
AOD : https://lpdaac.usgs.gov/products/mcd19a2v006/
Burned Area : https://lpdaac.usgs.gov/products/mcd64a1v006/
Mobility data : https://eodashboard.org/?indicator=GG&poi=GG-GG
geojson used in the app : https://data.ca.gov/en/dataset/california-counties2
geojson used in the data collection : https://www.census.gov/programs-surveys/geography/guidance/tiger-data-products-guide.html
#forestfires #air #human #activity #mobility #temperature #moisture #precipitation #no2 #co2 #aod #burned #fire #forest
This project has been submitted for consideration during the Judging process.