Final Report: Reducing Pesticide Application Rates by Elucidating the Pesticide Dose-Transfer Process

EPA Grant Number: R823100 (R821171)
Title: Reducing Pesticide Application Rates by Elucidating the Pesticide Dose-Transfer Process
Investigators: Hall, Franklin R. , Chapple, A. C. , Downer, Roger A. , Ebert, Timothy A. , Hislop, E. C. , Taylor, R.A. J. , Wolf, T. M.
Institution: The Ohio State University
EPA Project Officer: Hahn, Intaek
Project Period: July 1, 1995 through June 1, 1997
Project Amount: $202,691
RFA: Exploratory Research - Environmental Biology (1995) RFA Text |  Recipients Lists
Research Category: Biology/Life Sciences , Health , Ecosystems


Current estimates of application efficiency (Toxicant reaching target organism and resulting in mortality) range from about 1% for herbicides to less than 0.001% for insecticides. Modest improvements in application efficiency could significantly reduce pesticide use and environmental load. There is no general theory of the dose transfer process (atomization to biological effect - Fig 1 in Paper 1) from which efficacy can be predicted from basic physico-chemical parameters. Research at the Laboratory for Pest Control Application Technology (LPCAT) has been directed to understanding each stage of the dose transfer process with the objective of improving pesticide application efficiency. The primary thrust of this research was to understand the interactions between the target (or non-target) organism and the pesticide application. For insecticides applied to plant surfaces there must be some mechanism mediating the transfer of toxicant from the leaf surface to the insect which may result in mortality.

Laboratory research indicated that smaller droplets were more efficacious for insecticides. However, this relationship did not show up as a consistent trend in field data. While this could be the result of increased variability in the field, we hypothesized that it was due to the way research had been conducted in previous experiments. However, any new experimental design needs to have the potential to reconcile apparently contradictory results from laboratory and field data (Introduction in Paper 1).

The first step in developing this new design is to define deposit quality. Deposit quality is the physical arrangement on the surface of individual molecules of the toxicant. Individual molecules may be aggregated or dispersed to form more or less uniform coatings of toxicant. Since agricultural applications involving atomized fluids produce discrete clumps, areas of the surface where droplets have impacted and been retained versus areas where they have not, the absolutely uniform coating is not usually achieved. This is in contrast to deposit quantity (or total dose) which is the mass of toxicant per unit area.

The second step is to define the elements of deposit quality. There could be several definitions depending on what scale is used to measure deposit quality. For any insect-insecticide system, the molecular level is probably too fine a scale for this portion of the dose transfer process. We chose deposit size (size) and deposit number (number) as two of our variables because our spatial scale of interest is between 0.05 mm and 10 cm. Very small droplets which would have very high drift potential are about 0.05mm. Ten cm is close to the distance a small lepidopterous larva might move on its host plant during normal feeding activities ? some will, of course, move more and others much less. Given the total dose, deposit size, and the number of deposits, the remaining factor must be the toxicant per deposit. Since the toxicant per deposit is dependent on the toxicant per droplet we refer to this variable as concentration. These variables (size, number, concentration) combine to provide qualitative variation in total dose per unit area. We now have an experimental system where quantity is distinct from quality and in a form suitable for experimentation.


With a new experimental design, it was hard to predict the kinds of results one should expect. However, we expected results which would be internally consistent and explain the inconsistent results found within field tests, and between lab and field tests: namely that smaller droplets were more efficacious in the lab, but were only sometimes that way in the field. In addition we expected that:

  1. The most uniform deposit structure would not be the most efficacious (otherwise ULV applications should not work at all ? because ULV applications produce lots of highly toxic but very small deposits which create a less uniform deposit structure than a high volume application).

  2. Insect behavior has a strong influence on the dose transfer process because it influences the encounter probability and encounter rate between the insect and toxicant deposit.

Summary/Accomplishments (Outputs/Outcomes):

Our current view of the transfer of toxicant from the plant to the insect does reconcile the differences between laboratory and field performance results. Laboratory studies exchanged size for number (applied smaller droplets but applied more of them) in a very linear way. For example, they looked at the effects of droplet sizes of 1000, 500, 250, and 125 mm. While this appears to be a simple experiment, the results can be highly variable depending on what conditions of number (and possibly concentration) are used to balance the change in size, because the experimental area covered by this range in droplet sizes is a triangle with its terminus at the 1000 mm droplet size. As a result, any experiment with a set of decreasing droplet sizes is trying to estimate the entire response surface (Fig. 2 Paper 1, or Fig 1 in Paper 2) using a straight line. Simple examination of Fig 5 (Paper 1) or Fig 2 (Paper 2) will show that it is possible to draw a straight line with decreasing droplet size which will provide any conclusion desired: small droplets are better, or there is no effect, or small droplets are worse. The problem is in trying to estimate a two dimensional surface by sampling along a line. Thus we have accomplished our first goal which was to devise a theoretical model which reconciled the differences between laboratory and field results. A summary of nine conclusions from these experiments is as follows: 1) Deposit structure can play a major role in toxicant efficacy: mortality ranging from 10 to 90% for the same dose, but different presentation scenarios (Paper 1). This suggests that we could vastly improve our utilization of toxicants by manipulating how we present the target organisms with the toxin. 2) A uniform deposit structure is not the best if one is forced to limit the total quantity of toxicant applied (Paper 1). This confirms hypothesis 1. Since current application strategies promote the application of uniform deposits (e.g. spray to run-off), we might need to rethink our recommendations. Also, we might be able to reduce the environmental toxin load by applying toxins with a different strategy than previously recommended. 3) Uniform deposit structures promote acquisition of sub-lethal doses. Because sub-lethal doses are necessary for the development of resistance, this observation supplies another mechanism for the development of insecticide resistance (Paper 1). 4) Percent mortality and the level of crop damage are not necessarily highly correlated. This lack of correlation can be brought about through the effects of deposit quality changing the time specific survival rates of different individuals (Paper 1, 2). Most laboratory bioassays use some measure of percent mortality as an indication of toxicant performance. However, if the goal is to increase yields, then measuring percent mortality could give one a false estimate of toxicant performance. This may help explain some of the difference between lab and field performance of some toxicants. 5) The correlation between percent mortality and crop protection can be weaker for nibblers (IUPAC presentation). To fully utilize the effects of deposit structure on toxicant performance one needs to include some measure of insect behavior because the optimal deposit structure shifts slightly depending on if the target is a nibbler or chomper. Furthermore, a toxicant which modifies insect behavior will also change how the insect interacts with the toxicant. 6) Variability in the response to toxin quality is greater for nibblers (IUPAC presentation). This could be a problem if one is concerned with an individual patch of damage, versus the average level of damage ? the mean level versus the standard deviation. 7) Small numbers of large deposits are more effective against nibblers (IUPAC presentation). 8) Large numbers of smaller deposits are more effective against chompers (IUPAC presentation). 9) Chomping is more efficient at gathering biomass than nibbling, but for transient toxins, temporary switching to nibbling will balance feeding with toxin detoxification rates to ensure that some individuals survive (IUPAC presentation).

Conclusions 6, 7, and 8 all show the effect of behavior on the insect toxicant interaction, which addresses Hypothesis 2. Hypothesis 1 was shown in conclusion 2.

Unexpected results: Showing that perfectly uniform deposit structures are not the most efficacious is disturbing, but may help to explain why the literature continues to report that "increased coverage did not result in increased biological effect." Most bioassays on insecticide toxicity rely on coating leaves or vials and exposing insects. However, if the uniform deposit structure is not the most efficacious, then these assays could be underestimating the potential toxicity of a compound. Furthermore, the graphs in Fig. 4 (Paper 1) mostly do not have the classical probit-model shape. Rather mortality curves from different deposit structures form a continuous gradation from the classic s-shape curve to a straight line. This suggests that the way in which we perform our laboratory bioassays is determining the nature of our results. It doesn?t invalidate previous bioassays, but extends them to encompass the full range in deposit structure (Paper 3).

We also determined the following five results (Paper 3): 1) The effects of dose and concentration are confounded in standard dose-response bioassays. By its very name, a dose-response bioassay explores the effects of increasing dose on a target organism. However, such assays increase the dose by increasing the concentration of toxicant. Knowing that concentration influences diffusion rates, one does not know how much of the observed dose-response is due to increased concentration and how much due to increased exposure (dose). This type of confounding causes problems in the design and interpretation of experiments which rely on the independence of different factors in the model (independence is an assumption of most statistical models). 2) Dose and time effects are confounded. Increasing the dose decreases the time to a response. Increasing the time increases the dose. This makes interpretation of many toxicity tests difficult at best. Consider an experiment where one covers a leaf with 1, 10, 100, and 1000 ppm solutions of a toxicant, and then introduces an insect. Observe mortality at 24 hours, and run a probit analysis. How does one know how much of the observed effect is due to dose, to concentration, or due to a change in the time involved. It makes a difference if 100% mortality occurred at 1 hour in the 1000ppm treatment, but 23 hours in the 100 ppm treatment. However, by the design of the experiment, the research would never know the difference because he has not separated the effect of dose from the effect of time (or for that matter concentration) ? maybe think of it as increasing dose is compressing time. Possibly this distinction makes no difference in the field (we don't know), but as a conceptual model of the dose-transfer process it is very important to clearly distinguish the individual effects as well as the collective effect. Further examination of these factors may provide one or two key parameters which control the dose-transfer process and allow us to manipulate the system more effectively. 3) The use of a uniform deposit structure produces a graph best modeled using a probit or probit-like function. An equal number of deposit structures which are more heterogeneous could easily be modeled using standard least squares linear regression methods. As a conceptual model, the probit is no better (or worse) than a standard linear regression. The key to understanding the dose-transfer process is being able to model this change and its effects on mortality and plant damage. 4) Including toxicant decay by photolysis can introduce significant breaks in the dose-response function. With no decay at night, mortality is fairly high and constant. During the day, mortality rates decline due to decreasing toxicant levels. This is another source of error in evaluating field performance based on laboratory tests. The models from laboratory tests seldom take into account the effects of toxicant decay on bioefficacy. Laboratory models use a continuous function with no allowance for time dependent changes in the slope of the mortality function. Environmentally sensitive chemistries (including bioinsecticides like Bt) should show more pronounced time dependent changes in the slope of the mortality function. Thus we need to be careful in how we evaluate these pest control agents, and how we deploy them in the field. 5) Many current bioassays run the risk of either overestimating or underestimating the biological effect of a toxicant. Which error is made depends on how accurately laboratory tests mimic field encounter scenarios, and the duration of exposure. At short time intervals (hours to a few days) the uniform toxicant assay underestimates actual mortality. At long time intervals (a few days and longer), it may overestimate mortality. This has broader implications for how we evaluate the effect of toxicants on target and non-target organisms. When the laboratory presentation of a toxicant (to an insect, mouse, human, or fungus) does not match the encounter scenario in the field, we are more likely to make an error in estimating the toxicant?s biological impact. We currently have no model to predict the magnitude or direction of the error.


  1. Deposit quality is best modeled as a mixture of deposit size, deposit number, and toxicant per deposit which combine to yield a total dose.
  2. Uniform deposit structures are not the optimal form of toxicant presentation.
  3. Percent mortality and the level of damage are not always highly correlated.
  4. Changes in insect behavior modify both the rate at which toxicant in encountered, and the way in which deposit structure modifies dose-transfer efficiency.
  5. Bioassay methodologies utilizing uniform coatings target a very small portion of the deposit quality response surface.
  1. We have started to build a conceptual model of the dose transfer process. However, it has only begun to include the effects of behavior, and toxicant degradation. Fully integrating these effects into our conceptual model will strengthen the model and provide further insights which could lead to significant reductions in pesticide use (through improved application and strategy utilization). The issue is made more difficult because behavioral components are not really independent. If an insect spends more time resting then it must be spending less time feeding. Hence resting and feeding behaviors are confounded. Changing how we view behavior and integrating it into our dose-transfer model will be a major undertaking.
  2. We have considerable flexibility in how we apply toxicants. Different hardware (nozzles, tractor speed, pressure, etc..) and different chemistries (surfactants, emulsifiable concentrates, wetable powders, etc?) and now precision farming techniques (GPS, GIS systems) provide a wide array of possible application scenarios. Thinking of these parameters as ?dial in technology? where different settings on the dial represent different application strategies, we still don?t know what is the optimal setting on the dial. Consequently, we have the technology to make a change, and the impetus to make a change through legal, social, and economic pressure. However, we don't have a well integrated model for deciding what change to make. We are working on replacing the trial and error approach with a more scientifically based decision process for determining what to change and when. However, our current efforts are narrowly focused on a few systems. Expanding our efforts to fungicides and herbicides would improve (or prove) the generality of our conclusions.
  3. We have shown that different behaviors alter the dose - transfer process. This would suggest that there is the possibility that how we are applying toxicants influences the development in insecticide resistance. An enhanced modeling effort examining this possibility may provide an additional tool for managing resistance development in insect populations.
  4. In a much expanded view of agriculture these results may influence how we view transgenically applied toxicants (toxins?). It may also play a significant role in ecological studies which focus on plant - insect interactions. So long as toxins are not uniformly distributed throughout the plant, insect behavior has the opportunity to influence insect and plant fitness.
Major Contribution:

The significant impact of these results is their implications for toxicant bioassay methodology. Most current bioassay methodologies strive for uniformity of toxicant in the target?s environment. While this does eliminate much variability in the experimental results, it is this same variability which could play a major role in determining the biological response to the toxicant in the field. Taking into account this variability in the dose-transfer process may improve the current disparity between laboratory and field results. Precision farming promises "dial-in" capacity for applying pesticides. Identifying the proper delivery prescription for various pest scenarios is part of our approach to clarifying the optimal deposit structure. In order to utilize scarce resources in an economic and ecologically friendly manner, we must understand these complex and interdependent relationships.

These results also have a potentially wide impact for testing many toxicants. For example: mammalian toxicity is sometimes evaluated by oral doses, or injection, or dermal application. Do these tests actually mimic the way these organisms would encounter a toxicant in the field? Furthermore, how does one relate the exposure of a mouse to that of a human? If the way in which we perform these tests is not indicative of the way in which these toxicants are encountered, then we don?t have any real clue as to the actual risk these compounds pose to us or to the environment.

Journal Articles:

No journal articles submitted with this report: View all 10 publications for this project

Supplemental Keywords:

RFA, Scientific Discipline, Toxics, Ecosystem Protection/Environmental Exposure & Risk, exploratory research environmental biology, Environmental Chemistry, Ecosystem/Assessment/Indicators, Chemical Mixtures - Environmental Exposure & Risk, Ecosystem Protection, Chemistry, pesticides, Ecological Effects - Environmental Exposure & Risk, Ecological Effects - Human Health, Biology, Ecological Indicators, fate, pesticide exposure, computer simulation model, dose-response, deposit topology, insecticides, agrochemcial, dose transfer process, herbicides, cypermethrin, agriculture ecosystems

Progress and Final Reports:

Original Abstract
  • 1996