Digital image analysis estimates of biomass, carbon, and nitrogen uptake of winter cereal cover crops

Abstract

Over-wintering cereal cover crops contribute to greater cropping system sustainability by reducing risk of erosion, adding carbon (C), and sequestering inorganic nitrogen (N) between fall harvest and spring planting of annual crops. Carbon and N pools can be measured by manual sampling and laboratory analysis of roots and shoots. However, this is expensive and unscalable to field levels. This study aimed to (1) determine relationships between shoot biomass and total biomass, C and N uptake of winter wheat (Triticum aestivumL.), cereal rye (Secale cerealeL.), and triticale (XTriticosecaleWittm.); and (2) develop models to estimate shoot biomass using green cover (%) from digital images. Shoot and root biomass, C, and N uptake of cover cropped fields were determined by destructive sampling within rectangular quadrats (97.8 × 20.3 cm) in fall 2011 prior to snowfall (58 fields) and spring 2012 prior to termination (30 fields). In addition, shoot biomass and digital images (top view of crop canopy taken at shoulder height) were collected in fall 2013 and spring 2012–2014 (266 quadrats) using four different cameras (Canon, Nikon, Apple iPhone, Motorola Droid) for a total of 589 images. Four approaches were evaluated to estimate green cover: (i) estimation from “Canopeo” software; (ii) RGB color space with a set cutoff; (iii) the Lab color space with fixed adjustment of the “a” channel per camera; and (iv) the Otsu thresholding method applied to the “a” channel of the Lab color space. Shoot biomass ranged from 37 to 2657 kg DM ha−1 and was linearly correlated with total biomass (R2 = 0.99), total C (R2 = 0.99), and total N (R2 = 0.92), independent of species or season. The Otsu method best estimated shoot biomass from green cover (R2 = 0.66; green cover ranging from 10 to 70%). This method also reduced differences among cameras and the influence of lighting conditions on green cover estimations. Exponential models for shoot biomass estimation from green cover (%) determined with the Otsu method had the highestR2 for cereal rye (R2 = 0.79), followed by triticale (R2 = 0.54), and wheat (R2 = 0.52). We conclude these models, combined with linear equations to predict total biomass, C, and N uptake from shoot biomass can be used to estimate the same from imagery. Additional research is needed to validate models across more locations, different growth stages, and to expand across a larger range of green coverage.