Remote Sensing and GIS Integration

(9/25/2019 12:00:00 AM)

Abstract

Geographic Information Systems (GIS) are a new and blossoming concept and continue to grow in complexity and utility thanks in large part to the proceeding and continual development of Remote Sensing.  Remote Sensing plays a large role in the enhancement of any GIS, and in most cases, allows data to become much more relatable and useful for anyone.  A GIS receives much of the data for its built-in layers from Remote Sensing platforms such as satellites, radars and airplanes.  Passive sensors contribute to imagery and data for land cover mapping, change detection, snow monitoring, thermal changes and terrain modeling.  Active sensors contribute heavily to data for extremely accurate terrain models known as Digital Elevation Models (DEMs).  These large quantities of data can be geo-referenced and integrated into one large GIS, allowing a user to access a powerful amount of information at one time with relative ease.  And as remote sensing technology continues to increase in resolution and power, the data base will enlarge and increase the potential power of users of a Geographic Information System.

Remote Sensing

Remote Sensing is the process of obtaining information about land, water, or an object without any contact between the sensor and the subject of analysis.  Satellite and digital imagery play an important role in remote sensing and GIS.  There are two types of remote sensors, passive and active sensors.  Passive sensors measure reflected electromagnetic radiation from the sun or are emitted from the earth’s surface (photography and satellite imagery).  Active sensors emit their own electromagnetic radiation and measure the response (radar, sonar, and LIDAR).  See the figure below for an example.

Remote sensing has many advantages over traditional data collection.  It provides a regional view over a large area, it provides repetitive looks at the same area over time, sensors see a broader portion of the light spectrum than the human eye, it provides geo-referenced digital data that is used a geographic information system, it allows retrieval of data for regions difficult to reach such as oceans or hazardous terrain, and it can collect large amounts of data in a short period of time.

Resolution is a fundamental term involved with remote sensing and there are four different types: Spatial, Spectral, Temporal, and Radiometric.  The earth’s surface area is represented by pixels of an image which is known as spatial resolution.  A large area covered by a pixel is known as low spatial resolution and a small area covered by a pixel is known as high spatial resolution.  Spectral resolution is the ability to resolve spectral features and bands into their separate components.  More number of bands in a specified bandwidth means higher spectral resolution and vice versa.  The position, number, and width of spectral bands determines the degree to which individual targets can be discriminated in an image.  Temporal resolution is the frequency at which images are recorded or captured in a specified place on the earth’s surface.  The more frequently it is capture, the better or finer the temporal resolution is said to be.  For example, a sensor that captures an image of agricultural land twice a day has better temporal resolution than a sensor that only captures that same image once a week.  Radiometric resolution is the sensitivity of the sensor to the magnitude of the received electromagnetic energy.  The finer the radiometric resolution of a sensor, it is more sensitive in detecting small differences in reflected or emitted energy.

Some of the popular satellites include: NOAA Advanced Very High-Resolution Radiometer (AVHRR) with 1100-meter resolution, Geostationary Operational Environmental Satellites (GOES) with 700-meter resolution, Moderate Resolution Imaging Spectroradiometer (MODIS) with resolutions of 250, 500, and 1000 meters, Landsat Thematic Mapper with a resolution of 30 meters, IKONOS with a resolution of 1 and 4 meters, and Quickbird with a resolution of 0.6 meters.

Geographic Information System (GIS)

A geographic information system is a system designed to capture, store, manipulate, analyze, manage, and present spatial or geographic data.  Remote sensing provides numerous sources of spatial data for use in a GIS.  Some examples include: Imagery as base data for GIS mapping, Land use and land cover mapping, Biophysical phenomena, feature extraction from satellite imagery, surface elevation determination, and landscape change.  GIS layers such as roads and rivers can be easily seen from aerial and satellite photos, this information is digitized, separated into layers, and integrated into a GIS.  Features can also be automatically extracted using computer segmentation algorithms and the resulting data can be imported into a GIS for further display and analysis.
Geospatial Tools and Techniques

Abstract

Basic geoprocessing can be used in Geographic Information Systems to answer a wide variety of mapping questions. Geospatial modeling is used to predict values in locations or in different timeframes for which data are not directly available. Map algebra can be used to apply mathematical operations to multiple inputs for modeling. Inputs are often weighted to reflect or test the impact of different variables on a given model. Environmental carrying capacity is often a derived from multiple models representing each contributing factor to a system’s carrying capacity. Geospatial change analysis may also be accomplished with basic map algebra showing the difference between one or more inputs. Building a conceptual flowchart or process model along with custom scripting is often used to facilitate iterative modeling analyses allowing researchers and managers to test multiple scenarios for varying conditions.

Basic Geoprocessing

Many common mapping questions can be answered in a GIS with basic geoprocessing. Frequently used geoprocessing techniques include intersections of locations, linear features and areas to show spatial coincidence as well as proximity analyses such as buffering, nearest neighbor determinations. Additional basic geoprocessing techniques are used to combine, summarize or simplify geospatial data. Processes such as merge, union and dissolve may be used to understand data in different ways or viewpoints or to prepare data for inclusion in additional analysis.

Buffering, nearest neighbor and cluster analyses provide standard proximity analysis in a GIS. These operations help answer the basic question of “what’s around” and can dynamically show spatial relationships.

The Environmental Systems Research Institute (Esri) provides an overview and details on basic geoprocessing at http://pro.arcgis.com/en/pro-app/help/analysis/geoprocessing/basics/what-is-geoprocessing-.htm.

Geospatial Modeling

Modeling is a technique that predicts values and attributes for locations that do not otherwise have data. In other words, modeling can fill in the gaps in geospatial knowledge or predict values at future dates when conditions may change. Modeling is also used to predict future changes or conditions in space such as predicting where certain fish species may occur and how they travel through regions or areas. Many of the techniques outlined below can be performed using most GIS packages. Additionally, add-on software specifically aimed at geospatial modeling is often available. The Geospatial Modeling Environment is one such software add-on available for free at http://www.spatialecology.com/gme/.

Interpolation and Extrapolation

Interpolation and extrapolation are basic modeling techniques used to model geospatial data. Starting with known values from as many locations as can be measured, extrapolation allows for calculated guesses at values everywhere (Figure 2). Interpolation performs the same operation but is programmatically limited to the geospatial extent of the starting, known locations (i.e., interpolation does not calculate values far and wide across the globe!).

Figure 2. Starting with values and locations in the left diagram, a value is calculated for all locations within the area via interpolation.

Because extrapolation/interpolation provide values in all locations, results are best expressed as raster data. The size of the grid cells may be set by the GIS analyst, but most GIS software use standard algorithms to determine cell sizes based on the spacing of the initial inputs. It is important to note that interpolation/extrapolation should only be performed on datasets where a value for the parameter being measured can reasonably be expected at any location. For example, salinity values in an estuary can be measured at any place throughout estuary, therefore, it is acceptable to model values in all locations.

Map Algebra

Interpolation and extrapolation help fill in spatial gaps and are often used to develop input data layers for additional analysis. Map algebra is a technique used, for example, to add GIS layers--perhaps resulting from interpolation--together to derive a predictive result (Figure 3). Any mathematical operation can be used on the input layers. Depending on the phenomenon being modeled, mathematical formulas can be simple or complex. The section on Habitat Mapping and Modeling provides some examples of more complex map algebra operations.

Figure 3. Example of input layers of salinity, disease incidence and water quality combined to model predicted shrimp across an area.

Weighting Inputs

A common technique used in map algebra is weighting--applying a multiplier or other modifier to a layer to control the impact of that particular layer on the modeled results. For example, the salinity layer shown in Figure 3 may be known to be more important to shrimp yield than the other inputs and so may be multiplied by some factor before the overall model operation is performed. Different weights can be applied until the overall results conform with other known maps if available and related to the phenomenon being modeled.

Using map algebra, perhaps with weighted overlaying, “hotspots” -- areas that are important or high value or “suitable” based on the weighting and/or algorithm applied. A “suitability” map can be thought of as a map showing areas where something may be expected to occur. The term is often used to refer to habitat for a given species (e.g., areas suitable as deer habitat). Suitability mapping is essentially “hot spot” mapping. In the case of fish disease outbreaks, all the conditions that would allow for disease could be mapped and added together to show the result as a hotspot map. This would be a “suitability for disease” map (or perhaps better said as a “areas with high potential for disease” map).

Carrying Capacity

Geospatial modeling is also used in determining carrying capacity for habitats or ecosystems. “Carrying capacity” in this sense refers to population sizes in numbers of individuals (or possibly spatial extent in the case of plant species). The carrying capacity of an area is often determined by applying the methods such as hot spotting, weighted overlays, etc. to numerous factors or thematic conditions affecting carrying capacity and then combining these using map algebra. An extensive carrying capacity study may therefore consist of map algebra performed on model results which are derived from map algebra in the first place.

Mapping Change

If layer inputs include time variables, basic map algebra may be used to determine the difference between two more layers to show a “difference map” or, in essence, before and after conditions. Documenting change over time and space can then be used as a means for examining what drove the change and how those drivers might continue to cause change. Predicting change by examining how things have changed so far underpins basic research into climate and resulting environmental change.

Routing

Routing, or least-cost path modeling, is another powerful technique that can be used to predict, for example, the paths different fish species may travel. Using modeling as described, a least-cost data layer can be developed that assigns values to each grid cell representing the “cost” of moving into that cell. A starting point is selected, and the GIS process checks the value of each neighboring grid cell and advances the route to the cell with the lowest, or least cost, value. A simple least-cost model is shown in Figure 4. A routing model could also be based on previously modeled habitat suitability and ran such that the path moves to the neighboring cell with the most suitable habitat. Rules can be set where neighboring cells have equal values. For example, in the case of a tie, the path may choose the cell to the west or south. Routing may also be performed on a vector data layer and is often used to model the best routes for delivery vehicles in a road network.

Figure 4. Simple least-cost path determination in a raster data layer.

Model Building and Scripting

Geoprocessing and modeling are made easier with model building and scripting. “Model building” in this sense is the creation within a GIS of a process flow chart that conceptually maps the modeling process including data inputs. Model building and scripting allow for more easily repeated processing and analysis to test different inputs and scenarios (Figure 5). Scripting allows for customs algorithms or other data manipulations unique to a given scenario.

Figure 5. Example of geoprocessing model. Blue ovals represent input data layers. Green ovals represent output layers resulting from processes noted in the rectangles. In some cases, processes are custom scripts called and executed by the model.

VIFEP (USAID workshop )