Image analysis allows us to derive new understanding from existing data by creating analytic maps for insight and knowledge. These raster (cell-based) layers can be used to map and model virtually anything that happens across the earth’s surface, like agriculture, planning, hydrology, climate, wildlife habitats, and much more. The big idea in this chapter is that imagery data—with its cell-based data structure—allows for the systematic and controlled analysis of a vast array of phenomena across multiple layers.
ArcGIS provides an analytic platform that enables you to combine imagery with other kinds of geographic information within analytic models. It’s simple. GIS organizes information as geographic layers. Meanwhile, earth imagery scenes and sensor data are also accessible as layers. ArcGIS provides thousands of analytic operators that can derive statistical information, model movement and flow across your surfaces, help you to combine layers to find the most and least suitable areas for your activities, and much more.
Imagery provides a versatile information feed—a virtual fire hose of information to your GIS. In turn, ArcGIS has a number of spatial analysis operators that enable you to gain deeper insight into and understanding of your information. These analytic tools enable you to address virtually any kind of question, such as deriving the statistical signal from your data, examining a sequence of events through time, and forecasting and predicting them into the future. Spatial analysis entails identifying and deriving new information layers to help solve all kinds of problems, such as finding the right places to build, analyzing your business performance or where that new market may be hiding in plain sight, evaluating and managing your agricultural production, or monitoring and forecasting diseases.
Virtually any problem we face can gain from analytical insight provided by ArcGIS. And imagery is always a critical information source in your analytic work.
GIS and imagery analysis have come together and integrated only recently. And with the advent of cloud and enterprise server computing, modern computing systems are capable of analyzing massive volumes of image information. Limits in modeling have been significantly reduced, enabling you to model and analyze your information in deeper, more profound ways.
The use of imagery for GIS analysis is nothing new. Throughout the past few decades, imagery sources such as multispectral layers, digital elevation models, and digital orthophotos have provided an analytical foundation for modeling and feature extraction. Here are some common examples.
A cost surface is a raster grid in which each cell value represents the cost to travel through it. Cost surfaces can model things like the optimal path for a bushwhacking fire crew, predicting how a fire might spread, or predicting the travel preferences for how a mountain lion might move across its habitat range. In this map, green areas represent lower travel costs for the big cats in semirural Southern California.
A cost path calculates the least-cost travel path for traveling from one location to another. Costs can represent a number of criteria, including actual monetary expenditure, but more often are related to time and effort required to complete the journey. In this example, you can see the best path for cougars to travel between two of their core habitat areas.
The very nature of cell-based data makes it ideal for certain kinds of advanced analytics that can’t even be considered with vector data. The ecological land units (ELUs) project is one such example. Four global layers (bioclimate, landforms, rock type, and land cover) were overlaid and combined to create a single output surface that portrays a systematic division and classification of the global biosphere using ecological and physiographic land surface features to describe and characterize each land unit. “This map provides, for the first time, a web-based, GIS-ready, global ecophysiographic data product for land managers, scientists, conservationists, planners, and the public to use for global- and regional-scale landscape analysis and accounting,” said Roger Sayre from the USGS.
Imagery can be used to automate the classification and locations of land into specific categories, such as different types of land uses and land cover. These derived layers can then be used as basemaps and, more interestingly, in subsequent analyses. Classifying a series of images from different time periods also enables analysts to explore how a location is changing, whether from natural forces or human interventions.
This forest change analysis tool evaluates the total tree cover loss and number of active fires within the selected area of interest, and shows the results according to the various land cover classes. The Global Forest Watch change analysis tool uses spatial and temporal information to allow you to conduct your own investigation on forest cover change, current land cover, and legal classifications in your area of interest.
Image segmentation is defined as a process of partitioning an image into homogenous groups such that each region is homogenous. This map shows the impervious surface of each parcel after these surfaces have been segmented out using feature extraction analytics in ArcGIS. This is a classic segmentation application.
A common question that GIS analysis helps to solve is, where is the best place to put something? Suitability models are used for just that purpose—to find the ideal place to build or preserve, depending on the objective. The problems addressed can be wide-ranging: where to locate a new shopping center, plant a crop, preserve a marsh, develop a windmill, or place solar panels on building rooftops.
For example, the relevant criteria for siting a new park might include 1) a vacant parcel of land at least one acre in size; 2) proximity to the river; 3) a location not too close to an existing park; 4) an area with mature trees; and 5) a location near the homes and work of many people. ArcGIS can readily model suitability for parks and other sites using raster data and imagery. Here are some more examples.
NOAA has used land cover as a way to predict water quality using GIS analysis. For example, water quality is typically higher in the vicinity of forests and wetlands, and typically lower in regions with industrial facilities and large parking lots. This story map offers an excellent overview of the approach.
The state of Minnesota modeled solar potential for the whole state by deriving solar radiation and aspect from elevation, vegetation, and other critical raster and imagery layers. This enables citizens to perform a quick, high-level assessment of where solar power might be a practical alternative for their locations.
Viewshed analysis involves analyzing what is or isn’t visible from a given location based on distance, terrain, and even land cover. It is an operation that enables you to identify the locations from which a particular landmark is visible; for example, from which areas in a park can I see a river, or how many windmills are visible from the town square?
This viewshed analysis determined the visual impact of a wind farm with four large turbines in a study area in England. The visual impact of building a wind farm in urban or semiurbanized areas has the potential to create controversy in the community. Being able to show from where a turbine farm would be visible before one is built helps utilities mitigate reaction.
This interesting story map uses GIS visibility analysis to tell the fateful tale of the Battle of Gettysburg in the American Civil War. At the moment General Robert E. Lee (at the red eye) committed to engage with Union troops, he could only see the troops in the light areas; everything shaded in gray (the much greater part of the Union’s strength) was invisible to him at that moment. Historians using personal accounts, maps of the battle, and a basic elevation layer were able to unlock the mystery of why Lee may have committed to battle facing such poor odds.
Hydrology is the science concerned with the earth’s water, especially its movement in relation to land. Because water moves in response to gravity, the elevation of the earth’s surface can be used to model how water moves.
Flood-prone canyons pose a significant threat to recreational users in the semiarid western United States. The NOAA National Weather Service Forecast Office in San Diego has recognized the flash flood risks that exist and has implemented enhanced flash flood services for two flood-prone canyons. This story map details the methods used to create public awareness of the highest-risk areas in the Anza-Borrego Desert State Park.
Watershed analysis layers provide an estimate of flood frequency as one of six classes from none to very frequent. Click any spot on the map to get a readout of the flooding frequency. This 30-meter resolution layer covers most of the continental United States, including Alaska, Hawaii, Puerto Rico, the US Virgin Islands, and several Pacific Islands including Guam and Saipan.
Raster data can be single band or multiple band, with only a few unique pixel values or with a full range of values in the given pixel depth. And there are a number of ways to visualize raster data as multiband imagery, in 3D and as dynamic time-series maps. For example, when viewing color aerial photography, you are often viewing a three-band raster dataset with an RGB (red, green, blue) renderer applied by default.
Analytical hillshading computes surface illumination as a raster surface with values from 0 to 255 based on a given compass direction to the sun (azimuth) and a certain altitude above the horizon. Terrain modeling and visualization helps to bring other information layers to life as shown in this map of soils in the Panoche Hills of California, west of Fresno.
The Coastal and Marine Ecological Classification Standard provides a comprehensive framework for organizing information about coasts and oceans and their living systems. This four-dimensional time series map includes the physical, biological, and chemical properties that are collectively used to define coastal and marine ecosystems. When presented in 3D, the data forms a stack through which an analyst can drill.
Image analysis has evolved dramatically since the first Landsat was launched. Initially, the emphasis was on image processing to make the imagery interpretable; later, to extract features which were used to populate GIS databases. Now, much of the required technology is commonplace. The new emphasis is on processing the imagery in ways that enrich our understanding of the world so that we can better forecast and manage what is about to happen and get ahead of the curve. This is what we are trying to achieve in agriculture, forestry, environmental resource management, urban planning, traffic management, and even in fields like law enforcement.
Image analysis doesn’t evolve in a vacuum. It is influenced by related things that are evolving. Today and into the future we see increased computing power with the parallel capabilities of cloud computing; we see more imagery from more modalities with better resolution and more collection options; we have access to massive existing GIS data collections before we even collect the image; and we have new and innovative ways to perform analysis. So where does this perfect storm of progress take us?
Let’s take agriculture as an example. In the United States, the vast majority of growers register their fields and crops with the USDA. There are good soils maps and elevation models are available for the entire country. With NEXRAD there is rainfall data from ground-based radar that is collected for the continent every five minutes for cell sizes of less than a kilometer. Combined with other temperature data and daily solar illumination data, crop models can predict what the state of growth should be for every field. This allows for the use of multispectral imagery to validate and adjust these crop models whenever new imagery is collected, whether it be from satellites, aircraft, or drones. In fact, drones make the data very personal for the individual grower, offering better resolution and collection frequency. The analytical results show anomalies where the grower should take action to mitigate moisture issues, nutrient deficiencies, or weed and pest pressures. The net result is better understanding of production whether it is at the national level or for precision agriculture at the field level.
Image analytics are no longer just about making a pretty picture. Rather, they combine the science of remote sensing with all the other available sensor and GIS data to model the important processes that occur every day in our landscapes and affect our lives.
The goal of cartography or any style of information design is to highlight what is significant about the data. In many cases, when we let the data come to the surface, it is a sophisticated spatial analysis that becomes the map, or the information display.
When precipitation falls on the surface of the earth, much of it is captured in storage (such as in lakes, aquifers, soil moisture, snowpack, and vegetation, among others). Precipitation that exceeds the storage capacity of the landscape becomes runoff, which flows into river systems. In urban areas, pavement and other impervious surfaces drastically increase the amount of surface runoff, which sweeps trash and urban debris into waterways and increases pollution and severity of floods. In agricultural areas, surface and subsurface runoff can carry excess salts and nutrients, especially nitrogen and phosphorus.
Bathymetry is the study of underwater depths of lake beds or ocean floors. In other words, bathymetry is the underwater equivalent to topography. This map explores the world’s oceans and their bathymetric features.
When Hurricane Irene struck the Outer Banks of North Carolina in 2011, the storm surge and winds carved two new channels through Pea Island. The main transit route back to the mainland was destroyed. Lidar and imagery were flown by state and regional transportation agencies to collect multispectral data and surface information.
The damaged road was the only way in and out for local residents. Not only did the roadway itself need repair, but the surrounding beach also had to be rebuilt as a buffer zone to protect the new road. As soon as the imagery was flown and analyzed (mere days after the event), it was made available to responding agencies and proved invaluable in getting the infrastructure rebuilt.
The state of North Carolina deployed a simple app that allowed officials to begin making calculations about how many truckloads of sand would be required to replace all that had been washed away by the storm. By drawing different-sized shapes on the ground, they were able to provide some realistic estimates for how much sand was needed to get the roadway and beach repaired as rapidly as possible.
Many local governments use the amount of impervious surfaces to calculate the storm water bill for individual properties. Using dynamic image processing, impervious surface features are extracted from multispectral imagery and then used to compute the total square footage of impervious surface per parcel, as shown in this example from Charlotte, North Carolina. This analytical calculation provides an excellent illustration of the synergy that is enabled by the integration of GIS and image processing.
ArcGIS Spatial Analyst is an extension of ArcMap that augments the capabilities of ArcGIS Desktop by adding a range of raster spatial modeling and analysis tools. It is used to solve complex problems such as optimally locating new retail stores or determining the most promising areas for wildlife conservation efforts. While beyond the scope of this book, it’s an important tool in the serious analyst’s kit.
This course is for people who know something about data analysis and want to learn how the special capabilities of spatial data analysis provide deeper understanding. You’ll get free access to the full analytical capabilities of ArcGIS Online, Esri’s cloud-based GIS platform. Previous experience with GIS software is helpful but not necessary.
Ground surfaces that are impenetrable to water, known as impervious surfaces, present serious environmental problems. Runoff from storm water can cause flooding and carry contaminated materials into lakes and rivers. Due to these hazards, many local governments enact fees on land parcels with high amounts of impervious surfaces. Among these is the local government of Louisville, Kentucky. However, to place a storm water bill on properties, they need to know the area of impervious surfaces contained in each land parcel.
You’ll help them by calculating the impervious surfaces of a single Louisville neighborhood. With the aid of an ArcGIS Pro task, you’ll extract bands from a multispectral image of the neighborhood to emphasize urban features like roads and gray roofs. Then, you’ll segment and classify the image into land use types, which you can reclassify into either pervious or impervious surfaces. After assessing the accuracy of your classification, you’ll calculate the area of impervious surfaces per land parcel to supply Louisville with the necessary information to determine storm water fees.