ClearSky Week 3 — Big eye in the sky

Modis

Modis is a dataset provided by observation of 2 nasa satellites. One for day(Terra) and one for night(Aqua) two satellites collect data on the near-IR band and a couple of other features to globally and daily detect fire’s on 1km²,500 m² and 250 m² granules fitted over a sinusoidal earth projection.

This dataset is compiled by NASA and analysed by University of Maryland to give values to pixels and represent

• Cloud pixels
• Land pixels
• Water pixels
• Fire pixels

The dataset also includes coordinates for all of the pixels in the format (lat, lon). We plan to use this to integrate geographical information and weather information in conjuction with this.

Dataset is conveniantly formed in composite fire information on 8 day pixels and we plan to use the cnn to perform a time series analysis on this data.

Since it is just detection by satellite imagery and it is liable to misdetections when atmospheric conditions interfere with observance the dataset also includes a confidence interval for each fire pixel, this indicates how sure we are of the fire being located on that fire pixel. We plan to use only nominal/high confidence intervals since any misprediction means misuse of the resources which are valuable in a deadlock state.

The dataset combined with GFWED dataset of the nasa should as we propose provide enough prediction ability of the fire spread!

See you after a very busy week! :)

-Deniz Zağlı & Ahmet Şahin Yener