Drone-Based Remote Sensing Methods for Modeling, Mapping, and Monitoring Vegetation

A vernal pool reference site at the Elsie Gridley Mitigation Bank in Solano County. The right pane shows the drone-captured aerial photo. The center pane shows an elevation heat map based off of a digital surface model (DSM) created from the drone aerial photographs. Low elevations are shown in blue and green while higher elevations are shown in red. The left pane shows topographic contours in ¼-foot intervals that were derived from the DSM. This model will be used to analyze pool depth and compare that to hydroperiod and vegetation communities; the resulting information will then be used to design restored vernal pools with specific parameters based off these reference pools. Photo courtesy Airphrame.

A vernal pool reference site at the Elsie Gridley Mitigation Bank in Solano County. The right pane shows the drone-captured aerial photo. The center pane shows an elevation heat map based off of a digital surface model (DSM) created from the drone aerial photographs. Low elevations are shown in blue and green while higher elevations are shown in red. The left pane shows topographic contours in ¼-foot intervals that were derived from the DSM. This model will be used to analyze pool depth and compare that to hydroperiod and vegetation communities; the resulting information will then be used to design restored vernal pools with specific parameters based off these reference pools. Photo courtesy Airphrame.

Summer 2016 Ecesis, Volume 26, Issue 2

The explosion of commercially available, unmanned aerial vehicles (UAV; i.e., ‘drones’) in the marketplace provides a novel tool to the restoration ecologist. With both fixed-wing and rotary-style (“copter”) drones capable of covering hundreds of acres in a day, there are numerous applications. Drones can produce a number of very useful products that can increase the accuracy and efficiency of traditional field studies to support restoration projects. This article explores the differences in the two primary drone platforms, the various products they can produce, and a variety of applications pertinent to restoration ecology.

Drone Platforms

Drones come in two flavors, each with advantages and disadvantages: 
Fixed-wing aircraft are essentially small airplanes powered by a small electric or gas-powered motor. They require less maintenance than more complex rotary-style drones and are more efficient, thus they can cover more ground per flight. They can also carry heavier payloads, thus they can be paired with more advanced sensors with more capabilities. Fixed-wings do require more space for takeoff and landing, and they cannot hover or fly very slowly which is useful for some applications.

Rotary-style drones are probably more recognizable due to their popularity with enthusiasts. They typically have 4-8 propellers that allow the device to hover and fly slowly while maintaining stability in moderate winds. Disadvantages of this drone type include reduced flight times on a single battery which reduces the amount of ground they can cover, and reduced payload capacity which means they may not be able to carry all the sensors that fixed-wing aircraft can. 

For both systems, flights are planned and operated using tablet or smart phone applications that can run on iOS or Android operating systems. The flight area is input using a map interface and the application automatically sets up a flight path for the operator. Multiple flights may be needed to complete the entire flight depending on the size of the area and platform being used. In these cases, the drone returns to its original starting point so the battery can be replaced and then the unit is re-deployed to continue the data capture where it left off. 
Both systems also require certain certifications for use. The Federal Aviation Administration requires a Section 333 Exemption from companies or individuals that use drones for commercial use. In addition, drone operators must hold a valid pilot’s license. Currently the Section 333 Exemption process takes approximately 4–6 months to obtain. However, the federal government is currently revising the commercial drone operating requirements which should make the process less onerous. 

Drone Products

The two pieces of data most commonly produced by drones are photographs and topographic ground models. While these are not new to our industry, these data sets are indispensable — the drones allow for ultra-high quality images at a very low cost with near instantaneous deployment/data capture.

The primary product that drones produce for restoration purposes are high-resolution, georeferenced, orthorectified mosaic imagery. These are produced by taking hundreds of individual, vertical aerial photos that are mosaicked into a single aerial photo covering tens to hundreds or thousands of acres. Because drones fly at such a low elevation relative to traditional aerial imagery-capturing techniques (i.e., fixed-wing piloted aircraft), they are capable of producing photos with resolution quality of two centimeters per pixel. In comparison, most commercially available aerial photographs are of much lower quality, traditionally a 1–3 foot range. 

In addition, post-processing can provide a high-resolution digital surface model (DSM) to generate topographic contours. Since the individual photos taken from a drone overlap one another, stereophotogrammetry can be used to reconstruct a 3-D model of the scene. Ground control points are recommended in conjunction with this process to maintain a high level of accuracy since the typical GPS unit on a drone camera sensor has an accuracy of only 2–4 meters.

Specialized Products

One of the biggest limitations for drones has been payload capacity. However, with advances in technology, smaller sensors are now allowing more specialized equipment to be carried. One of these is a multi-spectral imaging camera. While traditional cameras take panchromatic photographs that represent reflective colors as the human eye perceives them (commonly called three-band imagery: red, green, and blue bandwidths), specialized equipment can obtain multispectral photographs that capture additional wavelengths not perceivable to the human eye (such as infrared or ultraviolet). The most common is four-band (red, green, blue, and near-infrared, or NIR) imaging which allows the production of color infrared photos which are very useful for vegetation studies. Although the technology exists to capture many more bands, the equipment is too heavy for most commercially available drones. 

Very recently, tiny LiDAR devices (acronym for Light Detection And Ranging, a surveying technology that measures distance by illuminating a target with a laser light) have hit the consumer market allowing a drone user to procure LiDAR-based elevation data. While expensive, LiDAR is much more accurate and higher resolution than DSMs produced from overlapping aerial photos. In addition, LiDAR can be used to capture both bare earth DSMs (elevations of the ground surface under vegetation) in addition to DSMs of the site that includes vegetation height (which is what is obtained from stereophotogrammetry). 


Mapping Vegetation Communities

One of the most basic applications for high-resolution drone aerial imagery is mapping vegetation communities. Mapping the extent of various communities — necessary for understanding the baseline conditions of a site or for monitoring the development of habitats following a restoration project — is a common practice for restoration ecologists. While commercially available satellite photos may have sufficient resolution for coarse vegetation classification, the high-resolution photos offered by drones often permit the ecologist to identify vegetation communities to a very detailed level. This is especially true if flights are possible over the site at various times of year, when the phenology of various species may be captured to assist in classification, or if a DSM is used to add topographic data which would include vegetation height data to the vegetation to further differentiate communities. Furthermore, multi-spectral imagery that offer NIR wavelengths can be especially useful in distinguishing between vegetation types. 

Remote Sensing

Remote sensing software can assist in classifying vegetation types in the aerial photograph. The software divides an image into polygons that contain pixels of a similar color, intensity, and texture. The user creates different categories (e.g., vegetation types), and then selects representative polygons of a specific vegetation community to ‘teach’ the program which category they belong to. Once this has been completed for each of the vegetation types in the image, the program classifies all of the remaining polygons into one of the user-defined vegetation communities. The user reviews the results and errors are reclassified and the recognition process is repeated until the error rate reaches a predetermined (i.e., acceptable) level. Adjacent cells of the same classification are merged and the resulting file is exported into GIS. Once in GIS, the files can be cleaned, acreages can be calculated, and maps can be generated.

Weed Mapping/Census

Using a similar process to vegetation community mapping, individual target weed species can also be identified and mapped. Examples include artichoke thistle, Arundo, perennial pepperweed, and Phragmites, although others are undoubtedly possible. For these readily identifiable species, the process is as accurate as field surveys for a fraction of the time (e.g. cost), especially for larger survey areas. For some species, such as Arundo and pepperweed, timing of the flight may be an important consideration. For Arundo, if it is underneath the canopy of deciduous riparian trees, it is beneficial to capture imagery during the winter when the canopy trees are leafless. For a smaller species such as pepperweed, photographing it when in bloom with its distinctive white top is helpful to pick it out from the surrounding species. 

Mapping Watercourses

In areas where vegetation coverage is sparse, such as those found in the deserts of the arid west, drone-derived DSMs can be a vital tool for mapping watercourses. Without the presence of vegetation to distort the true surface of the ground, GIS spatial analysis can be performed to identify flowlines and classify them into watercourses based on values from the DSM. By modeling watercourses ahead of time, field-mapping efforts are drastically reduced. This is extremely useful for areas subject to high temperatures or rugged terrain which can pose hazards to field staff. Aside from safety benefits, reduced field time translates into cost-savings for the client, which is always appreciated. 

Wetland Restoration Planning

When designing complicated wetland restoration projects, such as vernal pool complexes, it is helpful to have functional reference sites to use as prototypes for the design process. In vernal pools, pool depth is directly related to hydroperiod, which in turn drives plant species establishment and distribution. Minor fluctuations in topography of even a few inches can have a major effect on which plants grow in that area due to the changes in local hydrology. Therefore, having a detailed digital surface model of the pool topography is essential for understanding the distribution of topographic zones within a given pool. By capturing high-resolution imagery during the dry season when the pools are empty, an accurate DSM can be prepared for the area. A follow-up drone flight can be scheduled for the winter when the pools are full to obtain the water levels in the pool complexes. The two DSMs can be compared, while pool volume, depth, outlet elevation, and watershed can all be assessed. Individual pools can be analyzed and their vegetation communities assessed to determine target elevations and areas of restored pools. This general process has been used in the past but with using either expensive LiDAR data or more time-consuming ground-based data collection methods. With drones, the entire site can be captured at a fraction of the cost. 

Hydrology Monitoring

Wetland and stream hydrology is a common criteria used to assess restoration success. This is true for seasonal wetlands and vernal pools, streams, and tidal wetlands. The restored habitats are often compared to reference sites and are supposed to have similar hydrologic function (hydroperiod, depth, flow regime, etc.) as the reference sites. Data collection from drones can be used to assess at least some of these parameters. Geographical extent of ponding, or extent of high tide in tidal systems, is relatively easy to assess using aerial photographs. Similarly, by capturing aerial photos several times over the wet season, hydroperiod can also be determined. If a dry season DEM exists, GIS analysis can overlay a wet season DEM and determine depth of inundation. For large sites with multiple wetlands, drone data may replace many dozens of person-hours of repetitive hydrology monitoring of individual wetlands that may represent only a subset of the entire site. With drone-based aerial photographs, large sites can be flown in a matter hours providing a complete dataset of extent and depth of ponding for the entire site. This approach can provide a much more robust data set than traditionally available in a fraction of the time.

The Takeaway

While the specific types of products that drones are capable of providing may not be novel tools to the restoration ecologist, the ability to obtain them with drones make them indispensable for modern ecological monitoring and planning. With imminent changes in the legal framework for drone operation, these machines will be available to more operators. Simultaneously, technological advances will permit even more sensors to be reduced in size so they can be affixed to drones to capture even more data. Undoubtedly there are even more applications out there that are pertinent to your needs. We look forward to continuing to explore this exciting realm and hope to hear from others experimenting in this field about their success stories. — by Geoff Smick (President) and Sundaran Gillespie (GIS Analyst), WRA, Inc.