Iowa Derecho Damage

The 2020 Midwest Derecho that impacted Iowa and its surrounding states saw wind gusts up to 110-140 miles per hour across a 700-mile front, caused widespread damage to infrastructure and the environment, and generated upwards of $4 billion in losses across the state (PBS, 2020). While damage was experienced all across Iowa, the Iowa City / Cedar Rapids (ICR) corridor area in the central-eastern part of the state was particularly impacted, with hundreds of homes and businesses suffering catastrophic damage in Cedar Rapids (NOAA, 2020).

An application of remote sensing imagery using change detection methods fused with socio-demographic data in a GIS is explored, providing insight and synthesis of information in an easy-to-use map format. There is precedent in using satellite-based remote sensing imagery to detect tree damage. Chehata et al. (2014) explored the use of high-resolution imagery with object-based change detection methods to map damage to trees after a wind storm in France. Furukawa et al. (2020) used high-temporal revisit PlanetScope imagery of the Japanese coast along with conventional change detection methods to examine the impact of modern "cubesats" on environmental remote sensing change detection research. These studies varied in their methods, both both demonstrate the power of space-based remote sensing to provide a panoptic view of the environment and the ability of GIS to produce detailed analysis on of the data.

This StoryMap is a showcase for the power of GIS to map, model, and visualize the environmental damage caused by the derecho. It presents data from a variety of sources, both as static images and interactive maps, as well as a curated set of tools in a GIS that allow the user to drill down into the map to view the data at a variety of scales and contexts. GIS can and should be used to solve real-world problems, and this StoryMap seeks to demonstrate exactly that: provide a GIS that can be interacted with by the public at large, regardless of GIS experience or knowledge.

The damage across the state was extensive. While the level of destruction varied by location, the central-eastern region saw the highest peak winds and thus some of the most severe effects witnessed.

Cedar Rapids and some of the smaller communities nearby suffered extensively. Many people were displaced from their homes, forced to seek shelter elsewhere in the middle of a global pandemic. Even for those who made it the storm without damage, the power for several areas remained off for days in the hottest month of the year.

In order to understand the impact that this storm had on the community, we need to go further than simply observing the damage through pictures and memories. Because the damage was not concentrated to one area of the state or aspect of society, this Story Map will concentrate on one of the more severe examples: the impact on tree cover in the Iowa City / Cedar Rapids area.

Trees were particularly hard-hit due to the nature of the derecho, with the sustained winds and high-speed gusts causing downed limbs, foliage loss, and even the total collapse or uprooting of entire trees. The monetary value of the tree population in the Cedar Rapids area is estimated to be around $112 million, meaning that replacement efforts for lost trees could cost the city in the millions of dollars (Jordan, 2020). There have been academic studies into the costs and value of an urban tree.

Song et al. (2018) performed a meta-analysis of other papers that examined urban tree costs, providing an estimated mean annual benefit of $44.34 at a cost of $37.40 per tree, although the median values placed the benefits at $21.19 and costs at $25.07. While the actual costs will vary on location, tree type, and other factors, there is strong evidence that good urban forestry governance and policy provides tangible benefits for the community.

Municipal governments therefore often place great emphasis on their urban tree cover. Having an adequate amount of foliage can lead to positive results, including cleaner air quality, decreased summer air conditioning costs, and overall aesthetic value (Vogt et al., 2015). Many cities therefore conduct inventories of the trees on public property, primarily in parks and right-of-ways along roads. Cedar Rapids and Iowa City both have recent tree inventories that are available to the public through a web-based GIS:

These inventories provide detailed information about each tree in the public areas of town, but can take months to complete at the cost of hundreds of thousands of dollars. Additionally, they only capture the trees that are on public land, so many of the trees in the area are not included. In our exploration of the derecho damage, this means that municipal tree inventories are not suitable for creating a comprehensive tree damage assessment in a rapid manner. They can, however, serve as a baseline in a study such as this where we are looking to identify damage trends across a large area.

With our baseline of trees in the ICR area identified, we must now bring in our damage estimates. One of the simplest means of identifying changes in vegetative cover is to use a time series of vegetation indexes. In the slider map below, a Normalized Difference Vegetation Index (NDVI) is presented for two dates, one before and one after the derecho. Healthy vegetation is shown in green, while non-vegetative matter such as bare dirt, water, and roads will show up as red. The image on the left is from 6 August, while the image on the right is from 19 August, nine days after the derecho. A clear difference in vegetative health can be observed in the Iowa City metropolitan area of the images from roughly one week apart.

Quantifying and locating the damage using satellite imagery does has some challenges and limitations. While many trees were knocked straight over and were removed quickly, others were partially downed or for a variety of reasons left in place despite being irreparably injured. Other trees were affected in a way that will result in long-term degradation, but did not immediately show visible signs to this effect. All of this means that the imagery captured by satellites must be processed in a manner sensitive to different types and degrees of damage. To accomplish this, a technique called Change Vector Analysis (CVA) was used. Put simply, CVA identifies the "magnitude" and "direction" of changes in different spectral components of multitemporal images. Wang and Xu (2010) compared different change detection techniques, including CVA, using Landsat data to examine hurricane damage to forests. Their study showed that CVA was effective in detecting changes in forests, with a roughly 80% accuracy demonstrated.

CVA permits an more complex interpretation of the differences between image, moreso than simple NDVI time series. In the Google Earth Engine application below, a CVA methodology was coded to perform an analysis of NDVI and NDWI (Normalized Difference Water Index), two vegetation indexes highly correlated with vegetative health.

As can be seen in the CVA map, there is significant damage (in red) across the central and eastern part of the state, as well as sporadically in other locations. The CVA algorithm also detects the continued growth of crops around Iowa (in green), which helps highlight areas of significant crop damage, as these areas declined in vegetative health while other areas of the state improved. For damage to urban tree cover, the map is able to "drill down" into each of the urban areas in Iowa. Cedar Rapids shows significant signs of damage - in the city center, residential neighborhoods, and city parks. When compared to a town of similar size in the region, it is striking how much more vegetative change, as evidenced by NDVI and NDWI, Cedar Rapids experienced from the derecho. This is in line with the Cedar Rapids government's estimate that two thirds of the city's urban tree canopy was destroyed, and that one out of every three trees was in need of attention from an arborist.

The image to the left presents a snapshot of the damage estimate, tessellated into hexagons to help spatially normalize the view. It is important as a GIS professional to find ways that can help the intended audience comprehend the data being presented. Here, the damage was widely but non-uniformly distributed; a tessellated grid can help organize the information into a more digestible form, and the color map applied further accents the hotspots of the damage. Yet, while we can provide imagery products and polished analysis to suit, it is always better to provide people the tools so they can interact with the data directly.

The next section presents a curated set of data for use in convincing a group of officials, politicians, or other interested persons the value of a GIS in considering problems with a spatial component.

While the embedded maps in this story map are limited in the options that can be changed on-the-fly, they still allow the user to navigate and explore the data. Drill down into these examples of products generated using a GIS to explore the data at a variety of scales.

This first map is an interactive display of the estimated vegetation damage in Cedar Rapids, using Sentinel-2 Multispectral Imagery to perform Change Vector Analysis. The map is interactive, and because the the data for the imagery is at 10 meter resolution, as you zoom into different neighborhoods the map will resolve the damage estimate more precisely than seen at the current scale.

Since this was a weather event, let's bring in some weather data. The National Weather Service collected data from monitoring stations and created a KMZ with the footprint of the maximum wind gusts observed. We can convert that KMZ to a polygon feature class, clean up the data, add a color gradient, establish a legend, and plot it on the map. On top of that, we can take a MODIS-derived damage estimate from Google Earth Engine that has been extracted from a raster into a tessellated regular polygon grid to enable observation at scale, removing locations with no damage while accenting with different colors the locations with strongest damage. This map is navigable, and when you move in closer to the city level, the tessellated raster is replaced by the MODIS raster for finer detail.

Here's an example of how we can add in demographic data to help synthesize data from a variety of sources and types. There are two socio-economic layers in this map: one of residential homes retrieved from the Linn County Assessor, color coded by property value, and a second of income data from the US Census Bureau American Community Survey. Again, we use a restriction of the zoom display to optimize which layers appear at which scales, taking advantage of both a broad (ACS data) and detailed (parcel map) way to display income information, along with a broad (tessellated MODIS) and fine (Sentinel-2 raster) damage layer. Zoom in to view the detail, or zoom out to see the broad. This is useful in exploring the economic factors in conjunction with the spatial distribution of damage, all at a variety of scales.

These are three curated examples of interactive maps that we can easily put together into a single GIS for even greater fusion of data into information. Each example shows a different way we can bring in remote sensing imagery, analysis layers, or even ancillary data like weather maps or socio-economic details. The power of a GIS is the ability to take large amounts of spatially-complex data and present it in a format that is accessible to the user. These maps were deliberately designed to be understood by anyone, regardless of their familiarity with particular software or GIS in general.

Communities around the Midwest suffered a severe impact on August 10th. Homes rendered uninhabitable, damage to crops past time when they could be replanted for the season, trees snapped in half - all of this and more happened in the span of a few hours. Cedar Rapids, as the largest town in the epicenter of the storm, experienced a substantial amount of damage, in particular to its urban tree cover. It will take years of work and millions of dollars to restore the trees in the city; even just clearing out all of the downed limbs and debris is still ongoing five months after the derecho.

There are questions that a GIS can help answer. Was there a difference in spatial distribution of damage, such as to lower-income areas where perhaps tree maintenance had not been prioritized? What prioritization efforts should be made to provide the best response from the city, maximizing the efficiency of the assumedly limited arborist attention. Moving forward, how can the city can use assessments such as these to ensure equity in replanting efforts? It is important that data-driven analysis is used to provide decision makers with the clearest information possible so that they can best serve the community.

Being able to create maps such as these can help translate difficult or spatially-complex problems in to easily-digested products, allowing even non-GIS professionals the ability to interact with the data. Additionally, maps and broad-scale assessment tools are crucial in the aftermath of disasters, so preparing and prepositioning GIS assets in advance can help responders and workers better organize their resources. GIS and remote sensing have a great deal to offer the environmental sciences, and city officials should take advantage of their benefits in preparing for the future.