Precision Agriculture

Interrelationships of Insect Density and Physiological Variations in Upland Cotton

Status of cotton pest control in New Mexico - Traditionally (some 35 years ago) New Mexico had no primary pest insect problems in cotton. Pink bollworms Pectinophora gossypiella were present but cold winter temperatures generally kept populations low. As winter temperatures gradually warmed, pink bollworms became an economic problem and after the introduction of boll weevils Anthonomus grandis approximately 10 years ago, both boll weevils and pink bollworms required widespread insecticide applications.

We are currently are in the fourth year of, basically, a Malathion® boll weevil eradication program in New Mexico. Only four adult boll weevils were taken in traps in 2002. In addition, a referendum was passed in the spring of 2002 to eradicate pink bollworm basically using noninvasive pheromones and sterile males. Pink bollworm populations were drastically reduced during the growing season of 2002.

If these two initiatives continue to be successful, the only significant pests in cotton in New Mexico will be bollworm, Helicoverpa zea; lygus, Lygus hesperus; aphids; cotton aphid, Aphis gossypii, cowpea aphid, A. craccivora and green peach aphid, Myzus persicae and silverleaf whitefly, Bemisia argentifolii. Bollworms, lygus, whiteflies and aphids have traditionally been secondary pests in New Mexico. Their populations have usually been induced by insecticide applications. Lygus, bollworms, aphids and whiteflies are often controlled by the beneficial complex and may be further reduced by proper cotton management, i.e. avoiding luxury consumption of nitrogen and water (Leigh 1996).

Problem and Solution - Unfortunately there is no fast and reliable way of obtaining density estimates of insect populations to determine if the beneficial complex will control harmful species. Obtaining accurate insect population data is an extremely labor intensive, manual, and slow process. Consequently, growers resort to fast, often poor, sampling procedures or preventative maintenance insecticide applications based on primary consumer counts. In turn, beneficial insects are killed and resistance, production costs and environmental insults increased. There is a need to get to the point where simple routine sampling, management, and integrated biological control practices replace routine insecticide applications. Implementing these practices in the world's second largest pecan orchard, by New Mexico State University (NSMU) has reduced insecticide applications by some 90% (LaRock and Ellington 1996). The work proposed here will overcome this remaining hurtle and enable fast, accurate population estimates of the insect species that inhabit cotton fields and biological solutions to harmful insect problems.

Prior and Current Research

Complexity of Cotton Ecosystems - Cotton ecosystems are complex. In Arkansas about 600 species of predators representing 45 families of insects, nine families of spiders and four families of mites may be associated with cotton (Whitcomb and Bell 1964). In the San Joaquin Valley of California, 300 to 350 predacious and parasitic arthropod species may be found in cotton (van den Bosch and Hagen 1966). In New Mexico, some 15-30 insect genera routinely occur in large numbers from cotton field samples (Ellington et al. 1984a, Ellington et al. 1984c). Ellington et al. (1984d) identified over 50 species of herbivorous arthropods which feed on cotton in the Western United States and Ellington and Carrillo (2001) identified 125 species of parasitoids associated with cotton and alfalfa in New Mexico. Although these systems are complex, there are only 15-30 insect genera in New Mexico cotton that interact in an economic way.

Sampling Insects - The objective is to sample the primary consumers, parasitoids and predators in agricultural systems to predict mortality of phytophagous species. Cotton ecosystems are difficult to sample because there are so many species present and they generally occur in high numbers in clumped distributions (Ellington et al. 1996). Obtaining statistically relevant density estimations of primary consumers, predators and parasitoids requires large replicated samples (Ellington et al. 1996).

. Acquiring samples - A small four wheel drive (hydraulically driven) self propelled platform [Insectavac (IV)] with a 4,200 cfm high vacuum fan has been designed and built to take representative cotton insect samples (Ellington et al. 1984b).

. Calibration - The IV collector was calibrated for 24 genera of insects by comparing 83, 100 ft relative vacuum samples in three cotton fields to 830, 2.4 ft absolute samples in the same fields. Accuracy of density estimates depends on density and degree of clumping of insect populations and sampling protocol. Most primary consumers, predators and parasitoids can be collected very efficiently using the IV; however, bollworm eggs and larvae must be counted in traditional ways. IV densities can be converted to absolute or sweep net densities if needed (Ellington et al. 1984c).

. Sample size - The optimum size and number of samples needed to estimate the mean density of 24 genera of insects was determined by vacuum sampling 590, 100 ft quadrats end to end 12 times over a three year period in three cotton fields. Data for each date and location were pooled and the data analyzed as a nested design. It was found that it took four 100 ft vacuum samples per cotton field to estimate the mean density of most insect species to within 80 ± 5% of their true value (Ellington and Southward 1996). Density estimates for genera collected by sweep net were only within 7 ± 5% of their true value.

. Counting and Classifying Insects - Because it may take four people a week to manually count and classify the insects from one cotton field, a concerted effort has been made to automate the counting process over the years (Atmar et al. 1973, Gonzalez 1986, Ellington and Carrillo 1994, Waldie 1996, Morris 1994, Gassoumi et al. 1999). The most recent effort using soft computing techniques (fuzzy logic and neural networks) has yielded over 95% counting accuracy of 13 insect genera in the laboratory (Gassoumi et al. 1999 and Gassoumi et al. 2002). Although we have worked on the automatic counting and classification of insects for many years, the digital equipment and algorithms to carry out fast reliable identifications has only become available in the last two years. Further tests are needed to incorporate these new counting technologies into our system and confirm results at the field site.

Predicting Primary Consumer Densities

The complex group of arthropods found in agricultural ecosystems may be composed of host-specific and host-nonspecific primary consumer, parasitoid, predator and hyperparasitoid species. These arthropods may interact in positive, negative or neutral ways, depending on behavior and factors that influence natality, mortality and migration (Ellington et al. 1988a). Each system must be evaluated on its own merits.

. Migration - Alfalfa remains a primary source of beneficial insects in New Mexico. We have documented 120 species of parasitoids and two dozen predators which migrate from ¼ - ½ mile to adjacent cropping systems each time an alfalfa field is cut (Ellington and Carrillo 2001) (Fig. 2). Furthermore, the density of the beneficial complex in cotton in New Mexico returns to previous levels shortly after the residue from insecticide applications dissipate (Ozkock 1997), suggesting constant migration of the beneficial complex from alfalfa to other cropping ecosystems.
Figure 2. Average of four Insectavac insect samples taken 150, 300, 450, and 600 feet from the edge of a 360A cotton field down-wind from a 360A safflower field, Boswell farms, Corcoran CA.

. Predator Power - Some parasitoid species and most predaceous arthropods are relatively host-nonspecific. A host switching parasitoid or predator can stabilize an otherwise unstable host-parasitoid interaction (Murdock 1969).

Ellington et al. (1997) in an extensive six-year study based on over 1200 (100 ft) samples, taken with the IV vacuum sampler in 31 different cotton fields, found predators were associated with various primary consumers 163 times and various other predators 191 times, suggesting switching by predators may occur readily. If this is true, an invading primary consumer might be primarily controlled by non-host specific, switching predators. To initially test this idea, 5 bollworm eggs per plant were glued to 40 cotton plants (200 eggs) 5 times in 3 cotton fields in 1998 (10,000 eggs total per field). Egg mortality was evaluated daily for 2 days after initiation of the test. It was found that 50 and 85% bollworm egg mortality occurred one and two days after initiation of the test.

Final R2 values from regressions of total predators on bollworm egg mortality in these cotton fields was above 90%. Additional mortality might be expected from parasitoids and also third day exposure of eggs and first instar larvae to predators, parasitoids and possibly weather factors (Ellington and El Sokkari 1986). Thus we expect these mortality factors to be conservative. Our observations over the years suggest that native immigrating beneficial insects from alfalfa do stabilize phytophagous population outbreaks in cotton and that most of this mortality is from predators. Insecticide spraying based only on harmful insect monitoring fails to make use of this "free" mortality factor and contributes to expense, resistance and environmental pollution. Semistressing cotton through reduced applications of water and nitrogen along with the biological mortality factors already working in our cotton fields will stabilize lygus, bollworm, aphid and whitefly densities below their economic thresholds, leaving many fields in a state of biological control.