BUREAU OF LAND MANAGEMENT Ground-Based Image Collection and Analysis for Vegetation Monitoring Technical Note 454 October 2021 Suggested citation: Cox, S., D.T. Booth, and R. Berryman. 2021. Ground-Based Image Collection and Analysis for Vegetation Monitoring. Technical Note 454. U.S. Department of the Interior, Bureau of Land Management, National Operations Center, Denver, CO. Disclaimer: The mention of company names, trade names, or commercial products does not constitute endorsement or recommendation for use by the Federal Government. Production services provided by: Bureau of Land Management National Operations Center Information and Publishing Services Section P.O. Box 25047 Denver, CO 80225 BLM/OC/ST-22/001+6711 Ground-Based Image Collection and Analysis for Vegetation Monitoring Technical Note 454 Authors: Sam Cox Natural Resource Specialist Bureau of Land Management D. Terrance Booth (retired) Rangeland Scientist Agricultural Research Service Robert Berryman Software Developer U.S. Department of the Interior Bureau of Land Management October 2021 G N RI O T NI O M ON Acknowledgments TI A T E G E V The authors appreciate the valuable input from the following reviewers and collaborators: R O F S SI Tammie Adams, BLM National Operations Center Y L A N Marcell Astle, BLM Rawlins Field Office A D N Jim Cagney (retired), BLM Northwest District Office A N O Ken Henke (retired), BLM Wyoming State Office TI C E Douglas Johnson, Oregon State University L L O C Dusty Kavitz, BLM Buffalo Field Office E G A John Likins (retired), BLM Lander Field Office M D I E Neffra Matthews (retired), BLM National Operations Center S A B Cheryl Newberry, BLM Rawlins Field Office D- N U O R G ii TECHNICAL NOTE 454 G R O U N D -B A S E Contents D IM A G E C Abstract............................................................................................. iv O L L E Introduction..........................................................................................1 CT IO Part 1. Ground-Based Image Collection................................................................3 N A N 1.1 Use a high-resolution digital SLR camera.......................................................3 D A N 1.2 Protect the camera in the field.................................................................3 A L Y S 1.3 Acquire images with a 0.5 m2 field of view (FOV) ...............................................4 IS F O 1.4 Document the sampling location..............................................................8 R V E 1.5 Place a label in each image to link images with plot locations...................................8 GE T A 1.6 Adjust camera settings ........................................................................9 TIO N 1.7 Transect method.............................................................................13 M O N 1.8 Systematic grid or random plot method ......................................................14 IT O R 1.9 Alternative to the monopod: The “freehand” method ..........................................15 IN G 1.10 Download images and convert RAW files to JPG..............................................16 Part 2. Image Analysis ...............................................................................17 2.1 Manual image analysis with SamplePoint .....................................................17 Summary ...........................................................................................20 Appendix 1: Printable Plot Labels ....................................................................21 Appendix 2: Ground-Based Image Collection Checklist for Vegetation Monitoring .....................22 References ..........................................................................................23 TECHNICAL NOTE 454 iii G N RI O T NI O M ON Abstract TI A T E G E V Vegetation monitoring is integral to maintaining healthy and productive public lands. Virtually all R O activities on public lands have the potential to degrade native vegetation, with cascading effects of F S SI soil erosion, loss of resources, and diminishing ecosystem services. Monitoring vegetation allows land Y AL managers to recognize problems and implement management solutions in a timely manner to preserve N A resources and ecosystem benefits. This technical note describes ground-based image collection D N for vegetation monitoring, including best practices and equipment details. This technical note also A N O describes image analysis using SamplePoint software, which produces foliar cover measurements with TI C potential accuracy exceeding 90% (Booth et al. 2006). By acquiring and analyzing ground-based images, E L L land managers can monitor more area during a growing season, analyze imagery during the off-season, O C E improve statistical power through larger sample sizes, and maintain permanent records of resources. G A These benefits allow land managers to make more informed and defensible management decisions. M D I E S A B D- N U O R G iv TECHNICAL NOTE 454 G R O U N D -B A S E Introduction D IM A G E C Vegetation monitoring is integral to maintaining permanently monumented transect locations. O L healthy and productive public lands. The Bureau This method is congruent with traditional LE C T of Land Management uses various methods to sampling methods, such as line point intercept. IO N monitor vegetation, such as line point intercept It is not statistically robust since each transect is A N and ocular estimation. Digital images acquired considered the sampling unit—plots near each D A vertically (nadir) over study plots and field other are spatially autocorrelated and cannot N A L transects provide a permanent visual record of be considered independent samples (Stohlgren Y S IS the resource, suitable for numerous methods of et al. 1998; Goslee 2006). Transects are suitable F O analysis, both automated and manual. Ground- for measuring trend of representative areas but R V E based image monitoring can be completed faster are statistically insufficient for describing a large G E than other quantitative field methods, provides area of interest (Coulloudon et al. 1999). TA T managers with a permanent record of the resource IO N that can be reanalyzed or used for different • Sampling plots in a systematic grid over an M O N purposes later, and allows rangeland specialists to entire area of interest achieves a broad measure IT O divide their time efficiently by acquiring images of that area. Plots are located far apart and are R IN during the growing season and analyzing them thus not spatially autocorrelated to the degree G later in the year when vegetation is dormant and of plots along a transect. Since each image is unsuitable for field identification. an independent sample, this method results in higher sample sizes and is more statistically Unlike aerial orthoimagery, which emphasizes robust than the transect method. Though the continuous image coverage of an area, ground- net area of plots may be equal between the based image collection strategies emphasize transect and systematic grid (e.g., 20 plots of discontinuous, high-resolution, high-quality, 1 m2 each), high measurement precision is consistent sampling, so images reveal fine-scale achieved more efficiently with larger sample details, such as grass blades, pebbles, soil texture, size rather than larger plot size (Xiao et al. flower petals, and similar attributes that frequently 2005). Additionally, because a systematic grid facilitate feature classification to the level of is inherently spatially balanced, the results can functional group (forb, grass, litter) or species. be extrapolated across the entire area of survey. Derived measurements from these images are Traveling to all points in a systematic grid, treated as individual observations suitable for however, is more time consuming than sampling statistical analysis. with transects. Multiple methods and variations exist for ground- • The random plot method utilizes plots based image collection. The descriptions of three that are randomly placed within the area of recommended image collection methods are interest, often with a minimum buffer distance subsequently described, but numerous variations to improve spatial balance, resulting in less to these methods are equally valid.1 risk of systematic bias when measuring a heterogeneous vegetation distribution (Curran • The transect method involves acquiring images et al. 2020). Whysong and Miller (1987) reported along a field tape at 5-meter intervals, usually at that random plot placement was less susceptible 1 For detailed information about plots and transects, see Herrick et al. 2017. TECHNICAL NOTE 454 1 G to bias due to vegetation clumping and resulted within each image are secondary sampling units N RI in a lower Type I error (false positive) than and are also acquired using any of the three O T NI plots placed systematically, regardless of scale. methods previously described. This technical note O M Traveling to 20 random points typically involves contains two parts that describe each of these N O less distance than traveling to 20 systematic two sampling units. Part 1 provides a protocol TI A plots, resulting in time savings. for ground-based image collection (principal T E G sampling unit) to monitor vegetation, including E V R The steps described in this technical note equipment recommendations, instructions and O S F comprise a two-stage sampling design described tips for camera settings, sampling documentation, YSI by Elzinga et al. (1998). Each image is the principal data storage, and steps for the three image L A N sampling unit and is treated as an independent collection methods. Part 2 describes a method for A D sample acquired using any of the three methods measuring cover from images using SamplePoint N N A previously described (transect, systematic grid, software for manual pixel classification (secondary O TI or random plot). Individual sampled points sampling units). C E L L O C E G A M D I E S A B D- N U O R G 2 TECHNICAL NOTE 454 G R O U N D -B A S E D Part 1: Ground-Based Image Collection IM A G E C O Materials 4. Use of mobile devices, such as tablets or cell L L E • Monopod with hinged, quick-release mount phones, for image collection is discouraged. CT IO • Digital single lens reflex (SLR) camera with Although these will produce usable images, N A the basic monopod attachment, remote N zoom lens D shutter, camera controls, lens adjustments, A • Remote shutter release cable N A • One 100-meter field tape and two 2-foot rebar manual focus, and enhanced bit depth file LY S posts for transects capabilities (all subsequently described) IS F present in SLR cameras are either not present O R or not readily accessible in mobile devices. VE G Additionally, mobile device sensors are smaller, ET A 1.1 Use a high-resolution digital resulting in higher image noise given an equal TIO N SLR camera. resolution. Mobile device camera lenses are M O typically of lower quality than SLR camera N IT lenses, resulting in imagery of an inferior O 1. Use a high-resolution digital SLR camera that R quality. The quality difference is especially IN produces images with at least 20 million pixels G noticeable when an image is zoomed to (20 megapixels). The increased cost and file maximum, which occurs with photo sampling storage requirements of higher resolution during image analysis. Some advanced point- cameras are countered by the improved and-shoot cameras can be used successfully. analytical utility of resulting imagery. Although point-and-shoot cameras are less expensive, they do not provide the equivalent 2. Digital SLR cameras are more responsive to image quality provided by SLR cameras. For user inputs (e.g., no shutter lag) and usually these reasons, an SLR camera is recommended possess physical—as opposed to digital— for this application. shutter speed, aperture, focus, and ISO controls that can be readily accessed and adjusted in the field. This is a contrast to point-and-shoot 1.2 Protect the camera in the cameras that do not usually have buttons or dials for these settings, and the settings are field. Limit exposure to moisture, not as readily accessed through menus. sand, temperature extremes, and violent shocks. 3. Remote shutter release cables, which are necessary for the steps described in this technical note, are typically only compatible 1. Most SLR cameras can withstand a small with SLR cameras. Wireless shutter release amount of light moisture. A light drizzle or mechanisms can also be used and are required fog is usually not enough to penetrate the when using a long camera boom; however, environmental seals of the camera. Higher they are more expensive, require batteries, end SLR cameras typically have better weather and are more complicated to use. Infrared sealing than cheaper models. If a camera wireless remotes are unsuitable for field work will see heavy field use, look for models with because they require a direct line of sight to enhanced weather seals. When operating a the camera face. camera in such conditions, keep the lens glass TECHNICAL NOTE 454 3 G clean by keeping the camera pointing down 1.3 Acquire images with a N RI and wiping the lens frequently with a lens O 0.5 m2 field of view (FOV). Use T NI cloth. Even under a clear sky, be mindful that O a monopod with the camera M dew, mud, or dust may collect on the lens, N O necessitating cleaning. In all cases, keep the as high as practical above the TI A camera protected from the elements as much GET as possible. ground level.2 E V R O S F 2. Sand damages a camera through abrasive 1. Strive for the highest resolution imagery YSI action when the tiny grains come between possible. Image resolution is measured as L NA moving surfaces, such as buttons, selection ground sample distance (GSD), which is the A D wheels, and lens barrels. Do not set cameras distance of ground covered by one edge of a N N A down on sandy surfaces in the field, guard square image pixel. GSD should be ≤ 1 mm for TIO against dropping a camera in sandy soil, and accurate ground cover measurements (Booth EC avoid using a camera, if possible, in high wind and Cox 2011). L L O situations in sandy areas. If excessive sand is C GE encountered, use a soft dust brush and/or 2. Mount the camera to a monopod at a 20-30° A M canned air to clean the camera. angle so that when the camera is in a nadir D I E position, the monopod is 20-30° off-vertical S A B 3. Cold temperatures reduce camera battery life. and no part of the monopod is visible in the D- N Store batteries at room temperature. If it is image FOV (Figure 1). U RO cool outside, carry spare batteries in a pocket G close to your body. If camera batteries run low 3. Remove camera straps to prevent them from while sampling a transect, remove the battery hanging down into the image FOV. and hold it in your hand or pocket for several minutes to heat it up. This will usually allow a 4. Employ a bubble level, either by sticking it to short period of continued camera use. the back of the camera or some other parallel plane, to ensure the camera is oriented in a 4. For obvious reasons, avoid dropping nadir position. the camera. A hard-sided storage case is preferable to a canvas bag. To avoid 5. Image FOV should be consistent across sites, catastrophic lens damage, fit the lens with an users, and time. The FOV is determined by the ultraviolet (UV) filter. In addition to providing lens focal length and the camera height above beneficial filtering of UV radiation, the UV filter ground level (AGL). To reduce the FOV, either acts as inexpensive, expendable protection lower the camera height or increase the focal for an expensive lens. For remote field work, it length. Most SLR cameras are sold as kits with a is a good idea to have a spare UV filter in the zoom lens, but if your camera has a fixed focal storage case in case the first filter breaks, is length lens (e.g., 35 mm), then FOV can only be scratched, or is rendered too soiled for use. adjusted by altering the camera height AGL. 2 Modification of a method developed by Louhaichi et al. 2010. 4 TECHNICAL NOTE 454