SportsTurf provides current, practical and technical content on issues relevant to sports turf managers, including facilities managers. Most readers are athletic field managers from the professional level through parks and recreation, universities.
Issue link: https://read.dmtmag.com/i/251524
14 SportsTurf | February 2014 www.sportsturfonline.com to objectively quantify many turf quality parameters, including percent green cover, turf color (via a dark green color index, or DGCI), fertil- ity, chlorophyll index (i.e., "greenness"), and others. The objective na- ture eliminates variability associated with subjective visual ratings. In addition to their impact on visible light reflectance, many turf stresses largely impact reflectance in the near-infrared (NIR) region of the electromagnetic spectrum. Near-infrared is the portion of ra- diation just beyond that visible to the human eye, ranging from 700-1300 nm in wavelength. The NIR provides the ability to "see" stressed areas otherwise invisible. Near-infrared radiation can be de- tected and recorded using a modified digital camera. Modification costs are relatively inexpensive, costing about the same price of a new mid-grade digital camera; pre-modified digital cameras are also commercially available. Research at the University of Nebraska-Lincoln John Seaton Ander- son (JSA) Research Facility near Mead, NE, in 2010-12 has shown RGB and NIR information in digital images can be extracted with computer software and used to quantify turf quality and stress. Two commonly used agronomic measurements include chlorophyll index (CI) and the normalized difference vegetation index (NDVI). Al- though computed somewhat differently, each is an objective measure- ment of turf "greenness," calculated by mathematical manipulations of red and NIR reflectance data. Other methods based on analogous principles involve handheld sensors. Handheld sensors are commer- cially available that measure visible and NIR reflectance from turf and quantify a value. Researchers have demonstrated high correlations among multiple turfgrass qual- ity parameters with handheld CI and NDVI, making them robust, objective measurement tools. However, no attempts have been made to correlate these sensor data with a DIA system that incorporates NIR reflectance. A dual-camera (regular + NIR) DIA system may be a convenient, reliable, low-cost alterna- tive to handheld sensors for collecting turf qual- ity data. Regular and NIR-modified digital cameras used in tandem can record RGB and NIR reflectance data for each image. These data could provide CI and NDVI information, as well as percent cover, DGCI, and traditional DIA measurements. Furthermore, by combining DIA with drone technology, efficiency of collecting turf information increases dramatically. Drones provide the ability to image large areas, common in sports turf, in short time spans. For example, entire football fields can be imaged in minutes. By comparison, collecting imagery of equivalent area by hand would take several hours. Turf affected by various stresses, including water, fertility, disease, and insect damage, could easily be detected. In addition, be- cause drones can collect information on entire areas in one image, ef- fects of changing sunlight and cloud conditions are eliminated, increasing accuracy. Research conducted at UNL in 2012 investigated effectiveness of a drone-based, dual-camera (regular + NIR) DIA system for measuring CI and NDVI compared to handheld sensors. An ongoing deficit irri- gation field study established in 2009 was used. Deficit irrigation was applied via a linear gradient irrigation system, such that turf closest to the sprinkler line source received 80% evapotranspiration (well-wa- tered) and turf farthest received no irrigation (rain-fed); plots were di- vided into eight equal sub-plots that differed in irrigation and replicated four times. This design provided a broad range of turf quali- ties for analysis. Plots were mowed twice weekly at 2.5 inches, fertilized at 3 lbs N∙1000 ft-2∙y-1, and received regular pre- and postemergence herbicide applications. Aerial imagery was collected using a custom-built, GPS-controlled hexacopter equipped with a digital camera (Pixobot, LLC, Lincoln, NE). Aerial imagery of Bowie buffalograss (Buchloe dactyloide), 4-Sea- son Kentucky bluegrass (Poa pratensis), Apple GL perennial ryegrass (Lolium perenne), and Spyder tall fescue (Festuca arundinacea) was col- lected on 6 days approximately every 4 weeks from early April through late September. Imagery was collected in full sun between 1200 and 1400 hr. The NIR imagery was collected immediately following regu- lar image capture. A CI and NDVI were calculated for each image using the RGB and NIR data. The CI was calculated as (NIR / Red) – 1 and NDVI calculated as (NIR – Red)/(NIR + Red), based on equa- tions developed by previous researchers. Traditional DGCI (which does not use NIR) values were also calculated for comparison against CI and NDVI. Chlorophyll index and NDVI data were also collected using hand- held sensors. The CI and NDVI were measured using a Spectrum Technologies FieldScout CM 1000 chlorophyll meter and FieldScout TCM 500 NDVI turf color meter, respectively. Scores were averages of three random measurements taken in the center of each plot. Hand- held sensor data were collected the same days as aerial imagery. Our results showed strong correlations between drone-based CI and NDVI and handheld sensor data (Table 1). On average, drone-based CI data were highly correlated (R ≈ 0.84) with handheld CI values across turfgrasses. Similarly, drone-based NDVI values were highly cor- related (R ≈ 0.79) with handheld NDVI values across turfgrasses. The Field Science Turfgrass Handheld CI vs: Handheld NDVI vs: Drone-CI Drone-DGCI Drone-NDVI Drone-DGCI Buffalograss 0.78 0.75 0.71 0.71 Kentucky bluegrass 0.87 0.80 0.80 0.79 Perennial ryegrass 0.84 0.73 0.82 0.72 Tall fescue 0.87 0.74 0.82 0.75 Correlations of handheld chlorophyll index (CI) and normalized difference vegetation index (NDVI) sensors among drone-based CI, -NDVI, and -dark green color index (DGCI). (n = 184 each; all results were statistically significant at the 0.001 level)