Spatial Distribution of Broadleaf Weeds in North Carolina Soybean (Glycine max) Fields

Abstract

Spatial distribution of broadleaf weeds within 14 North Carolina soybean fields was characterized by fitting negative binomial distributions to frequency distributions of weed counts in each field. In most cases, the data could be represented by a negative binomial distribution. Estimated values of the parameter K of this distribution were small, often less than one, indicating a high degree of patchiness. The data also indicated that the population as a whole was patchy. Counts of individual species were positively correlated with each other in some fields and total weed count could be represented by a negative binomial for 12 of the 14 fields.

A Review on Remote Sensing of Weeds in Agriculture

“…In a study of 14 soybean (Glycine max (L.) Merr.) fields in North Carolina, Wiles et al (1992) found that up to 29% of each field's experimental quadrants, measuring 9.1 m by the width of the row spacing, contained no weeds. Using 35 mm aerial photography, Thornton et al (1990) concluded that only 18% of a wheat (Triticum aestivum L.) field was infested with wild oats (Avena fatua L.).…”

“…However, for aggregated weed populations, assumptions of equal sample variance do not hold, because the sampling locations are spatially related (Cardina et al, 1995). Later field investigations used a grid-based sampling approach to describe the spatial distribution of weed growth (Marshall, 1988;Mortensen et al, 1993;Wiles et al, 1992). Results showed that weed distributions could generally be characterized with the negative binomial function.…”

“…This meant that weeds grew in definite clumps and that the presence of a weed at a certain location increased the probability that another weed existed a short distance away. This was true for both broadleaf weeds (Wiles et al, 1992) and grasses (Marshall, 1988). Incorporation of the inherent patchiness of weed infestations into statistical models requiring weed distribution inputs then became a priority for much research in weed science (Brian and Cousens, 1990;Maxwell and Ghersa, 1992;Navas, 1991;Thornton et al, 1990;VanGroenendael, 1988).…”

“…Remote sensing has been widely explored as a tool for detection and mapping of weeds in agricultural crops (Lamb and Brown, 2001;Moran et al, 1997;Zwiggelaar, 1998). Supporting this practice, numerous inquiries into the nature of weed spatial distribution have confirmed that weeds generally clustered together in patches within a field (Marshall, 1988;Rew et al, 1996;Wiles et al, 1992). In addition, substantial portions of an agricultural field may contain absolutely no weeds or the existence of weeds below threshold levels at which the crop is threatened (Chancellor and Goronea, 1994;Johnson et al, 1995;Thornton et al, 1990).…”

An on-farm approach to quantify yield variation and to derive decision rules for site-specific weed management

“…In the context of the German reduction program for chemical plant protection, herbicide use needs to be controlled strictly in future and reduced to that considered absolutely necessary (BMVEL 2005). It is generally known that the distribution of weed seedlings in arable fields is heterogeneous (Marshall 1988; Thornton et al 1990;Wiles et al 1992;Mortensen et al 1993;Cardina et al 1995 andJohnson et al 1996). Uniform applications of herbicides are still the standard method of weed control.…”