The problem the technology or innovationaims to address

Thefirst step before engaging in any rangeland rehabilitation and managementproject should be centred toward mapping and assessing rangeland condition.Yet, ecologists and managers have been challenged to develop cost-effectivemethods of measuring changes in vegetation that are reliable and repeatable.Several studies have characterised natural vegetation at national or regionalscales. However, these studies require qualified staff and substantialfinancial resources, meaning exhaustive assessments cannot be conducted. Thereis a need for reliable rapid mapping and monitoring techniques for assessingrangeland condition, and estimating their productivity at the farm/communitylevel. User-friendly tools that can rapidly assess the occurrence and causes ofland-degradation processes have to be developed and field tested.

The technology/innovation

Withadvances in technology, large areas can be rapidly mapped and monitored.Combining information from low-altitude sensors and geospatial data appears tooffer an optimal path for developing a practical system for cost-effective,databased, rangeland monitoring and assessment. If an inventory of rangeresources is established, priority areas with the greatest potential forimprovement can be identified. In collaboration with Oregon State University,the International Center for Agricultural Research in the Dry Areas (ICARDA)Rangeland Ecology & Management Unit (REMU) has been developing quantitativemethods for estimating cover using globally positioned digital imaging coupledwith image analysis for several years. The project has developed software andprotocols that are repeatable and technician independent. These protocols use acontinuously recording global positioning system (GPS) device, a digitalcamera, and a computer to acquire and manage information.

Target areas

Theproposed technique would be disseminated to all countries in sub-SaharanAfrica.

Project details

Synopsis

Thisdocument describes a photographic monitoring protocol and software program thatcan be used to quantify vegetative cover in quadrats.

Summary

One ofthe most important indicators of rangeland condition and health is thepercentage of the soil that is covered and protected from raindrops by plantsand litter. Sequential measurements of cover at seasonal or yearly intervalscan indicate range trend. Unfortunately, plant foliar and litter cover isgenerally not measured, but rather estimated by technicians using the quadratmethod because quantitative measurement via intercept or point methods istedious and time consuming. Differences in experience and judgment lead tosubstantial differences in cover estimates between technicians and from onesampling period to the next. The project has been developing quantitativemethods for estimating cover using globally positioned digital imaging coupledwith image analysis for several years and have developed software and protocolsthat are repeatable and technician independent. These protocols use acontinuously recording global positioning system (GPS) device, a digitalcamera, and a computer to acquire and manage information. Specialised softwarelinks an image of the ground to a specific location, rotates and scales theimage and processes it into meaningful classes such as foliar cover, litter andbare ground. Original images, processed images, track logs and photo locationsare stored on the computer or a DVD for reference and archiving. Because bothpositioned and processed images are tagged with GPS coordinates, they can beviewed in a geographical information system (GIS).

Objectives

The maingoal of this project is to develop, test and disseminate algorithms andprotocols within a software package which can be used to measure per cent coverof foliage, litter and bare ground or other parameters of interest inelectronically positioned and defined quadrats. The specific objectives wereto:

1)    Rapidly classify ground-levelphotographs into meaningful groups based on colour.

2)    Save classification parameters so theycan be applied to other images taken under similar conditions.

3)    Automatically classify all photographsselected or all photographs in a directory folder.

4)    Export images as either classified ASCIIRaster maps or bitmaps with world files and projection information so they canbe saved as GIS data layers.

5)    Output summary tables with results ofthe classification for each image.

Materials and methods

Theproposed technique for measuring cover requires a staff-mounted digital camerathat is positioned to take a digital photograph vertically downward from afixed height when the platform holding the camera is level (Annex Figure 3.1)It consists of a Bogen Manfrotto 676B Monopod and a Bogen Manfrotto 3025/056 3DJunior Head to which a 5 by 8 inch (13 by 20 cm) platform has been mounted(Booth et al. 2004, Johnson et al. 2008). Attached to the platform is a bubblelevel, magnetic compass, continuously recording GPS unit, and the digitalcamera. The staff upon which the camera head is mounted can be either a pole ofa fixed height or a variable height monopod, as long as the height does notchange during a photo shoot (Annex Figure 3.1).

AnnexFigure 3.1: Apparatus used to obtain digital images of quadrats.

Figure3.2: Acquisition of digital images in the field using a staff-mounted cameraand a continuously recording GPS unit.

Image processing using VegMeasure software

VegMeasureis a computerised vegetation measurement program. It is the result of work wasdone on a series of projects by the Department of Rangeland Ecology andManagement at Oregon State University (USA) that measured and monitoredvegetation on rangelands, agronomic fields and riparian areas. This softwarepackage was developed to chart or map vegetation in a sample quadrat on theground, using photographs taken vertically downward. Special emphasis was givento accurately represent the position and area of plants in the quadrat anddocument their condition. Previously, hand charting of vegetation in quadrats,a very time consuming and tedious method, was used to collect cover estimatesin the field. VegMeasure is a successful attempt to speed up the process. Inaddition, the protocol presented helps standardise the measurement process sothat results are more uniform and reflective of the condition of vegetationgrowing on a site.

Sinceits beginnings, VegMeasure has also been used to quantify streamside vegetationusing low-altitude aerial photographs, over-storey cover using verticallyupward photos from ground level, and distribution of weeds from aerialphotographs by capturing distinctive flower colour. The software can be used onpermanent quadrats to document change through time, such as spring plant growthor reduction of leaf area by grazing. Charting vegetation across several yearswith VegMeasure has allowed the project to quantify the increase or decrease ofspecific bunchgrass and forb cover to measure growth. Individual plantpersistence and behaviour can also be measured via VegMeasure in vegetativecommunities (Annex Figure 3.3).

VegMeasureuses information gathered either from a digital camera, a video, a web cam, ora scan from conventional colour photographs or colour infrared photographs.These images can be processed in either real-time in the field using a laptopcomputer or they can be post-processed in the office. Most of the processinguses digital conventional colour images taken in the field with a digitalcamera. Typically, one should GPS the location of each image then process thedata later in the laboratory.

Given aset of images, VegMeasure analyse them according to either a preselectedalgorithm or a classification algorithm chosen by the user. The program thenproduces a report indicating the percentage of each class in each image as wellas all images collectively. At its simplest, VegMeasure will classify an imageinto two classes and show the results on screen where the user can evaluate thedata. Also included is a K-means algorithm that divides the image into a seriesof classes. This should facilitate identification of distinctive vegetation inthe image.

Additionally,VegMeasure supports interfacing to GPS units when they are attached via a USBport. When connected, VegMeasure automatically records the location (inlatitude and longitude) of an acquired image.

Literature on the technique

BoothDT, Cox SE, Louhaichi M and Johnson DE. 2004. Lightweight camera stand forclose-to-earth remote sensing. Journal of Range Management 57:675–678.

BoothDT, Cox S. and Johnson DE. 2005. Detection-threshold calibration and otherfactors influencing digital measurements of ground cover. Journal of RangelandEcology and Management 58:598–604.

BormanMM, Louhaichi M, Johnson DE, Krueger WC, Karow RS and Thomas DR. 2002. Yieldmapping to document goose-grazing impacts on winter wheat. Agronomy Journal94:1087–1093.

BradyWW, Mitchell JE, Bonham CD and Cook JW. 1995. Assessing the power of thepoint-line transect to monitor changes in plant basal cover. Journal of RangeManagement 48:187–190.

BråkenhielmS, and Quighong L. 1995. Comparison of field methods in vegetation monitoring.Water Air and Soil Pollution 79:75–87.

ClelandTM. 1921. A practical description of the Munsell color system, with suggestionsfor its use. Munsell Color Company, Boston, Massachusetts, USA.

Floyd DAand Anderson JE. 1987. A comparison of three methods for estimating plantcover. Journal of Ecology 75:221–228.

JohnsonDE, Louhaichi M and Vulfson M. 2003. VegMeasure: A C++ computer program forfield measurement of vegetative cover. In: ASPRS Proceedings, 2003 AnnualConference: Technology: Converging at the Top of the World. Anchorage, Alaska.

JohnsonDE, Louhaichi M, Vulfson M and Harris NR. 2004. VegMeasure user’s manual,Version 1.6. Department of Rangeland Resources, Oregon State University,Corvallis, Oregon USA.

JohnsonKD, Johnson MD and Louhaichi M. 2007. GeoAlbum user’s guide. Global GeomaticSolutions, Corvallis, Oregon, USA.

JohnsonMD, Johnson DE Louhaichi M, Clark PE, Wörz AL, Ndzieze SK and Johnson DE. 2009.VegMeasure 2 user’s guide. Department of Rangeland Ecology and Management,Oregon State University, Corvallis, Oregon, USA.

JohnsonM, Louhaichi M, Harris NR, Wörz AL and Johnson DE. 2008. A protocol formonitoring vegetation, bare ground and litter in scaled globally-positionedground-level imagery. American Society of Photogrammetry and Remote SensingProceedings, 2008 Annual Conference. Bridging the Horizons—New frontiers ingeospatial collaboration. Portland, Oregon, USA.

LouhaichiM, Johnson MD, Woerz AL, Jasra AW and Johnson DE. 2010. Digital charting techniquefor monitoring rangeland vegetation cover at local scale. International Journalof Agriculture and Biology 12:406–410.

LouhaichiM, Borman MM and Johnson DE 2001. Spatially located platform and aerialphotography for documentation of grazing impacts on wheat. GeocartoInternational, 16(1):63-68. 223

MacQueenJB. 1967. Some methods for classification and analysis of multivariateobservations. In: p 281–297, Proceedings of 5th Berkeley Symposium onMathematical Statistics and Probability. 1. University of California Press.