Tim Kuhn
GEOG 579B

Shortly before starting here at the University of Maryland, I accepted a contract position at the Federal Emergency Management Agency (FEMA) as a GIS Specialist. I’m currently working mainly on cartographic projects mapping logistics and shipping procedures throughout all fifty states and the accompanying territories in preparation for natural disasters. The skills and tools I started to learn in the area of remote sensing through this class will expand the scope of my work and make me a much more effective emergency management professional as I grow further into this career path, hopefully shifting toward disaster response from disaster preparation, where I am currently focused.

The first moments and days following a major disaster are pivotal in effectively responding to a crisis. The remote sensing tools that we learned in this class can help us gain a much better situation awareness after an event in a much quicker time. Applying photogrammetric techniques to both aerial and satellite imagery saves responders from having to survey the area on foot. It grants them the ability to gain a wider and more holistic understanding of the aftermath of catastrophic events.

I saw some of these techniques in action through my colleagues. It was when the Camp wildfires were ravaging the California landscape in November of last year. My co-workers were called to California’s Office of Emergency Management to offer the agency remote sensing support in tracking the spread of the fires.

 During that time, they invoked the International Charter of Satellites in order to collect imagery and to track and monitor the quick spread of the burn area. Our company, New Light Technologies, had recently unveiled a new web application on behalf of FEMA called the Wildfire Incident Journal (see figure below), which my team updated daily with real time location data of the burn area. When the flame was contained, our team is also involved in damage assessment. They took aerial footage of the town of Paradise, CA, which was especially hard hit during the disaster. Using visual and digital analysis of the imagery, along with 3-D modeling, they catalogued the devastation from the fires so that FEMA could begin the recovery efforts.

 The analysis and classification of digital multi spectral scanner data is incredibly useful for almost any type of disaster that the country may see. It was from hearing these stories that I found particular relevance in this class when we began studying the process of mapping burn scars. Having seen the process in a very tangential way during wildfire season last year, the lessons in the course provided great context for the work that was going on under the hood during that incident.  I now know the different types of sources of satellite data that are available for this type of work, whether it be from a Landsat, MODIS, or AVHRR, depending on the resolution needs of the moment. The different indices that we learned can be leveraged to study the landscapes that were affected by different incidents as well. In the case of wildfires, the Normalized Difference Vegetation Index (NDVI) — (NIR – Red) / (NIR + Red) — can be used to show the differences in vegetation levels before and after an incident. The Normalized Burn Ratio (NBR) — (SWIR – NIR) / (SWIR + NIR) — can show the actual land affected by the burn.

Emergency management also demands a working knowledge of a myriad of interpretation skills, including photogrammetry and both supervised and unsupervised classification of satellite imagery. It’s just as important that the user be able to analyze and interpret aerial photographs as it is for them to be able to train data and run analyses digitally. Geographic phenomena look quite different from above them they do on the ground.

The analyst must be able to intuit with the naked eye what features are based on the size, shape, shadow, toning color, pattern, texture, and situation of image components when doing a quick study of a region. Depending on the severity of a disaster, It might only require visual interpretation to determine the hardest hit parts of a city or county.

 However, there are plenty of times when a more nuanced investigation is required. This is where both supervised and unsupervised classification would be needed. To effectively utilize unsupervised classification, it helps to understand the different operations that are able to be performed on ENVI, as the user does not define any of the training areas used to define pixelation. The results come back from this automated automated can mean very different things, depending on whether K-means, ISODATA, or a hybrid classification approach is used. In supervised image classification, the user designates training areas that will guide the pixel determination in the process. The classification approaches that fall under this umbrella include minimum distance, parallelpiped, maximum likelihood, and  decision trees. Each carries with it different degrees and aspects of uncertainty, meaning there is interpretation necessary on the back-end of this procedure to verify accuracy through comparison to ground truth data via a confusion matrix.

Remote sensing is an entire process with multiple stages, each of which contains an abundance of nuance within each of its distinct parts — from the technology and systems used for data collection to data manipulation and interpretation. I’m excited to grow my understanding of every part therein as I continue forward in my nascent career in emergency management. It’s an industry where the stakes always remain high and timing is critical. While we face an uncertain future due to the escalating nature of weather events due to climate change, remote sensing stands to be an invaluable tool, one that can save countless lives in the years to come.