We also used multiple correlation analyses to determine which variables-plant (height, % nitrogen), soil (pH, electrical conductivity) or landscape (latitude, longitude, altitude, block, nitrogen fertilizer level, and irrigation level) had the greatest impact on aphid density, bollworm square damage, and boll weevil boll damage.  Results indicated that aphid density was significantly correlated with soil pH (r = -0.228), elevation (r = -0.125), and irrigation level (r = -0.278) and that the overall R² value was very low (0.13) indicating that these factors explained only about 13% of the variation in aphid density in cotton.  Bollworm square damage was significantly correlated with pH (r = -0.116), electrical conductivity (r = 0.229), latitude (r = 0.191), plant height (r= 0.316), block (r = 0.24), and irrigation level (r = -0.194) with an overall R² value of 0.221 indicating that the factors examined explained about 22% of the variation in bollworm damage.  Finally, boll weevil boll damage was significantly correlated with pH (r = -0.325), electrical conductivity (r = 0.235), latitude (r = 0.116), longitude (r = -0.242) plant height (r = 0.176), nitrogen fertilizer level (r = -0.129), block (r = 0.132), and irrigation level (r = -0.666) with an overall R² value of 0.73 indicating that the factors examined explained about 73% of the variation in boll weevil damage.

Finally, we examined the distribution patterns of aphids, bollworm damaged squares, and boll weevil damaged bolls using different sized sampling units to determine whether insect clumping would allow for the use of insecticides in only portions of the field.  For aphids we found that if 150 m by 24 row areas (blocks) were examined to determine if the aphid density was above the spray threshold (> 50/leaf), insecticides would only need to be applied to 50% of the field.  If smaller areas, 50 m by 24 rows, are examined than only 38% would be sprayed, while if even smaller areas (25 m by 12 rows) are examined the difference becomes small (37%) indicating that 50 m by 24 row areas are the best size for aphid management units.  This is somewhat different for bollworm damage, for 50 m by 24 row areas only 50% of the field would need to be sprayed; for the 25 m by 24 row areas 45% would be sprayed, and for the smallest area 30% would need to be sprayed.  Boll weevil damage was similar to bollworm, using the large areas, 100% would need to be sprayed, if the medium sized areas were used as management units then 88% would need to be sprayed, while if the small areas were used then only 58% would need to be sprayed.  For bollworm and boll weevil, the smallest area would be the best size for the management unit.

In conclusion, our findings indicate that variable irrigation and nitrogen fertilizer used in precision agriculture will affect insect pest densities and damage to cotton fruit.  Also, because nitrogen fertilizer and irrigation levels affect pest densities, both can be used in a model for predicting pest densities on a large scale.  However, on a smaller scale, pest densities were only slightly correlated with plant, soil, and landscape factors indicating that these factors alone will not be useful for predicting pest densities.  Finally, the clumped distribution pattern of insects and insect damage indicates that it should be possible to apply pesticides to only portions of a field, thus allowing the survival of natural enemies and the slowing of insecticide resistance evolution in unsprayed areas.

Lubbock: ANOVA results for lint, seed, and carpel weights were similar. Therefore, we'll just discuss lint and seed weights (per 2 m-row treatment plots) rather than repeating the same information for all three.  Both irrigation level and damage type (real or simulated) had significant effects on yield.  Yield was greater at the high irrigation level (means = 56.8 mg lint /plot and 90.9 mg seed /plot) than at the low irrigation level (means = 37.8 mg lint /plot and 58.2 mg seed /plot).  Yield was also greater for real damage (means = 48.8 mg lint /plot and 76.9 mg seed /plot) than for simulated damage (means = 45.8 mg lint /plot and 72.3 mg seed /plot).  The latter was probably due to high bollworm mortality in real damage plots.  This caused yield to be greater in real damage plots than in simulated damage plots, because we didn't take bollworm mortality into account when simulating bollworm damage levels at the three densities.  This was done to ensure that we obtained some information on damage even if high mortality in the real damage plots caused treatment densities to be indistinguishable.

There were significant interactions between irrigation level, and density and between irrigation level and infestation period (Fig. 6).  Neither density nor infestation period significantly affected yield at the high irrigation level (Fig. 7).  However, at the low irrigation level, yield decreased with increasing bollworm density (Fig. 8).  Also, yield was significantly less for the last infestation period (boll maturation) than at either of the earlier infestation periods (Fig. 8).

We have 4 main conclusions from the Lubbock bollworm fruit consumption-injury study.  First and most simply increasing the level of irrigation increases yield.  Second, simulated bollworm damage effects yield significantly more than real bollworm damage when bollworm mortality is not included in simulations.  Third and most importantly, increasing bollworm density decreases yield when irrigation is low but has no effect when irrigation is high.  Finally, bollworm damage that occurs during the boll maturation period decreases yield (in comparison with earlier cotton developmental periods) when irrigation is low, but not when irrigation is high.