Submitted to: HortScience
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 2, 2003
Publication Date: March 15, 2004
Citation: OLMSTEAD, M.A., WAMPLE, R.L., GREENE, S.L., TARARA, J.M. NON-DESTRUCTIVE MEASUREMENT OF VEGETATIVE COVER USING DIGITAL IMAGE ANALYSIS. HORTSCIENCE. 2004. V. 39(1). p. 55-59. Interpretive Summary: The traditional method of judging the proportion of the soil surface in a crop field that is covered with crop plants and/or weeds is for a farmer of crop consultant simply to look at the field or at a sample area. The downfalls of this traditional practice are that it is highly subjective, varying between observers, and varying for a single individual with his experience. Furthermore, the human eye can be 'tricked' into overestimating the presence of weeds when just few spots of greenery are present. Likewise, the eye tends to mistakenly underestimate weed presence when they cover most of the field, because the eye focuses on the small areas of bare soil instead. This error can affect the amount and rate of herbicide a farmer might apply to a field, or whether the farmer might decide to re-seed a young cover crop. Both are expensive practices that can have environmental impacts. Thus, it is important to develop and test objective methods of estimating the amount of ground covered by weeds and crop plants. Digital images of 1 m^2 sample plots can be analyzed by PC software that the user trains to distinguish between green and soil, and between the green shades of weeds and crop plants. Digital image analysis is accurate and highly repeatable. With the decline in cost of digital photography and PC software, this method should now be available not only to crop consultants but also to interested farmers.
Technical Abstract: Traditionally, vegetative cover has been subjectively assessed visually. However, visual assessment is thought to overestimate percent vegetative cover. Thus, a repeatable method to objectively quantify percent cover is desirable. In two vineyards near Prosser, WA, the percentage of ground surface covered by up to 15 different cover crops was assessed both visually and by computer-assisted digital image analysis. Quadrats in the cover crop were photographed digitally and the images analyzed with commercially available software. Areas of green vegetation in each image were identified and measured. Weeds in some images were differentiated from the cover crop by user-defined thresholds. Subjective visual estimates of percent vegetative cover were generally higher than those digitally estimated. Values for the visual estimates ranged from 5% to 70% in 1998 (mean = 52.4%) and 7.5% to 55% in 1999 (mean = 30.7%), compared to digital readings ranging from 0.5% to 24% (mean = 11.1%) and 10.3% to 36.6% cover (mean = 20.1%), respectively. The visual assessments had lower coefficients of variability in 1998 (CV 28.1) than the digital image analysis (CV 52.3), but in 1999, the values for the two techniques were similar (CV 41.2 vs. CV 45.7). Despite initial variations between the two methods, digital image analysis for measuring percentage vegetative cover is superior due to its accuracy.