Wild Fire Remote Sensing Plus Modelling

The NASA Earth Observatory has an interesting piece showing the practical consequences of the resolution of remote sensed data. They compare an image of a complex of wild fires last month taken by the older Moderate Resolution Imaging Spectroradiometer (MODIS) instrument with the newer Visible Infrared Imaging Radiometer Suite (VIIRS). The former has a resolution (in the IR band) of 1000 meters per pixel, the latter resolves 375 meters per pixel.

The article presents the equivalent images, along with a cool interactive “comparison” option. What you see is that the higher resolution imagery spots quite a few smaller and cooler fires than seen by the older instrument. This higher resolution is useful for fire fighters and for scientists trying to model and predict the course of wild fires.

Creating such models is hardly a simple matter (no matter what Hollywood might show you). A key method for developing models is to use data from the past, feed it to your simulation, and compare the computed results to what really happened. This often is a humbling exercise.

The article points to an example of such work, published by Janice Coen and Wilfrid Schroeder, “The High Park fire: Coupled weather-wildland fire model simulation of a windstorm-driven wildfire in Colorado” [1].

First of all, they created a coupled model, i.e., they adapted a code that models the atmosphere, specifically, as wind blows through mountains, and a second code that models wild fires. The coupling takes consideration of how the wind affects the fire, and how the smoke and heat of the fire affects the winds. In other words, in a coupled simulation, the sum is grater than the sum of the two parts.

These models are then fed data based on measurements of the state of the vegetation and measured winds, and run to calculate how the fires will spread (up hill, down wind) over time. The extent of the burn is compared to estimates at the time from satellites, aircraft, and ground observations.

If this sounds like a lot of work, it is. It is also skilled work—these pieces did not just clip together, they had to be very carefully understood to make sure that all the pieces are valid. See the article for details (ask your friendly librarian to help you get a copy).

Coen and Shroeder discuss the importance of high resolution models, because the wind was extremely turbulent 1-2 km above the ground, and varied quite a bit due to the rugged terrain. Fire, to, is a fine grained and fast changing phenomenon. Successfully modeling a wild fire required computations to the varying scale down to as fine as 100 meters horizontally, 10 meters vertically, and 1 second of time. This is a massive amount of data, and the whole enterprise is impossible if data is only available in big chunks (i.e., low resolution measurements).

Even with this heroic effort, “telescoping from the synoptic scale environment down to convective scale terrain-induced flow effects, the ignition and growth of the High Park fire in fuel conditions reflecting the widespread drought, and the weather-fire dynamic interactions during its first growth period” (p. 143), the model still did not completely match nature (which, of course, operates at “infinite” granularity).

They note a number of factors that could not be accounted for because data is not available (e.g., details of winds in the area of the fire, local details of the vegetation), and the lack of resolution vertically, and in the terrain model.

This study is great work, and is an excellent example of how computation and remote sensing are being used to understand our planet.

I do want to point out a couple of points.

First of all, consider the deep learning and practical intelligence that is required for such work. This is not a matter of clicking on some data, mashing up a couple of apps, and “poof” there is a simulation of the Earth. Both Hollywood and Silicon Valley imagine that computation is simple and obvious—it isn’t. It is very hard work, and takes a ton of knowledge.

This is especially important to remember when enjoying remote sensing imagery. No matter how pretty the images, drawing valid scientific conclusions from them requires a lot of extremely careful work. You can’t just eyeball it.

Second, politicians and pundits are full of it when they spout slander about how scientists are corrupt and rig the data to provide politically desired. Have a gander at a study like this and tell me that these folks are motivated by anything other than a desire to get things right. And they work hard to get it right. I have very little tolerance for such know-nothing denial of science.


  1. Coen, Janice L and Wilfrid Schroeder, The High Park fire: Coupled weather-wildland fire model simulation of a windstorm-driven wildfire in Colorado. Journal of Geophysical Research. Atmospheres, 120 (1):131-148, 2015.


Space Saturday

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s