Page 17 - Transitioning Turfgrass

th
6 ETS Field Days
05

A New Hyperspectral Based System

for the Estimation of Weeds and

Botanical Composition of Turfgrasses

Cristina Pornaro, Department of Agronomy, Food, Natural Resources, Animals, and Environment, University
of Padova (Italy), Loris Vescovo, Sustainable Ecosystems and Bioresources Department, Research and
Innovation Centre, Fondazione Edmund Mach, Trento (Italy), Michele Dalponte, Sustainable Ecosystems
and Bioresources Department, Research and Innovation Centre, Fondazione Edmund Mach, Trento (Italy),
Damiano Gianelle, Sustainable Ecosystems and Bioresources Department, Research and Innovation Centre,
Fondazione Edmund Mach, Trento (Italy), Stefano Macolino, Department of Agronomy, Food, Natural
Resources, Animals, and Environment, University of Padova (Italy)

The botanical evolution of mixtures and weeds (Macolino et al., 2014), or the combination of
invasion and their diffusion are important aspects the two (Leasure, 1949). A less used alternative
in turfgrass research. Monitoring plant population is plant count of single species, but this method is
dynamics in turf mixtures have shown increasing usually used in pot trials (Bailey et al., 2013; Ear-
interest by turf specialists in consequence of herbi- lywine et al., 2010) or in field trials using profile
cide-use reduction and the increase of low-main- sampler (Brede and Duich 1984a).
tenance management. Mixing two, three or more A number of studies have looked at identification
species is the simplest and probably best known of species from hyperspectral reflectance data-
method to reduce maintenance inputs for main- sets (Cho et al., 2010; Ghasemloo et al., 2011).
taining turfgrasses, especially in transition zones The use of hyperspectral image analysis for veg-
(Dunn and Diesburg, 2004). Mixtures of differ- etation mapping is mainly used for ecosystem
ent turfgrass species have better visual quality monitoring and remote sensing of vegetation
and reduced environmental and pest stresses (Malenovsky et al., 2009; Drusch et al., 2017).
compared with monostands (Salehi and Khosh- Several studies were conducted to classify tree
Khui, 2004). However, vegetation dynamics of species from hyperspectral imaging data with
different species determine uniformity and com- different results (e.g., Xiao et al., 2004) accord-
position of the turf, and are influenced by several ing to their spectral or spatial resolution (e.g.,
factors such as seeding rates (Brede and Duich Dalponte et al., 2009), and the adopted plat-
1984a), mowing height and frequency (Brede form (airborne or satellite; Vyas et al., 2011; Xu
and Duich 1984b), and fertilisation (Gough et al., et al., 2011). Specific studies were conducted on
2000). Indeed, studies on botanical composition species classification in grasslands (Cushnahan et
of weed invaded turfgrasses are very important al., 2016; Monteiro et al., 2008). However, the
especially with the recent rules in banning herbi- spectral variability showed to be very high and
cides imposed by the European Union. The most linked to the phenological stage and stress level
used methods for determining turfgrass species of the canopy (Cushnahan et al., 2016).
composition are visual estimation (Cropper et al., Hyperspectral sensors collect the light reflected
2017; Knot et al., 2017), point quadrat method from objects in a series of contiguous bands. The

13