| dc.description.abstract | This study considered runoff of pesticides from cotton fields using a rainfall simulator. The
Australian cotton industry is based on clay soils on low sloping land and uses a hill-furrow surface
geometry. These conditions are unlike those in many studies of pesticide dissipation, sorption and
runoff and there has been little previous research into hydrology, erosion and pesticide runoff in the
industry. Of particular interest was to characterise runoff of different pesticides, clarity the underlying
factors controlling pesticide runoff, and investigate management practices to reduce runoff of
pesticides with contrasting chemical properties, for the conditions found in the cotton industry.
Runoff behaviour of different pesticides has often been studied independently. Separating the
inherent behaviour of the pesticide and the conditions of the study, and comparing pesticides, is
difficult. Runoff concentrations depend on the hydrology and erosion of each site, and particularly on
the timing of runoff events after spraying. These factors can be controlled and/or measured using a
rainfall simulator. Furthermore, multi-residue pesticide analysis now allows study of a number of
pesticides simultaneously, so that their behaviour can be compared directly for the same site
conditions and management options.
Review of the literature indicated that a large variation in pesticide runoff is related to
application rate, formulation and placement, and dissipation, and that only a shallow soil surface layer
contributes pesticides to runoff. In this study, the analysis was simplified by only considering
placement on the soil surface (i.e. not on plant foliage) of liquid/emulsified formulations and soilincorporated
sprays. A simple conceptual framework was used to compare and integrate data from
simulated rainfall studies of pesticides used in the cotton industry and contrast them with data in the
literature. By comparing pesticide runoff to the concentration in the soil surface at the start of rain,
one set of factors, those that occur before the rainfall event - application and dissipation, were
separated from those that occur during the rainfall event - leaching and runoff extraction.
To complete the picture of how pesticides get from the spray nozzle to the edge of a field in
runoff, four main areas were considered - dissipation, runoff extraction, sediment-water partitioning
and management. Dissipation data were collected at four sites across the cotton growing areas and
runoff data at three of these sites, on soils with medium to high clay content. Some 14 pesticides were
studied, including the insecticide endosulfan (a- and (3-isomers and the breakdown product endosulfan
sulfate) at all sites.
Dissipation studies concentrated on the 3-6 weeks after application, when concentrations are
highest, and on the soil surface layer that contributes pesticides to runoff. For practical reasons, soil
concentrations were measured in the 0-25mm soil depth. While the major emphasis was on
dissipation of endosulfan from soil and crop residues, several organophosphate (OP) insecticides (chlorpyrifos, dimethoate and profenofos) and a range of herbicides (fluometuron, metolachlor,
prometryn, diuron, pendimethalin, pyrithiobac sodium) were also studied.
Dissipation in the 0-25mm soil depth followed first-order exponential decay, with one phase, for
most pesticides. However, large initial losses occurred for several of the insecticides studied: 35-55%
for endosulfan, but only when applied at higher temperatures, most likely due to volatilisation. Initial
losses of 50-75% occurred for the OP’s, dimethoate (inconsistently), chlorpyrifos and profenofos.
Dissipation half-lives in the 0-25mm soil depth (after initial losses) were 6-20 days for
endosulfan (total of a-, (3- and sulfate), 8-13 days for organophosphates, and 13-32 days for herbicides,
a-endosulfan consistently dissipated more rapidly than p-endosulfan, but the two isomers were
affected differently by site and/or climatic factors. Dissipation of endosulfan was similar for ULV and
EC formulations, for bare and covered soil and for band and blanket sprays, but was somewhat slower
after two applications. Only small amounts of endosulfan sulfate formed in dry soils, while more
formed in temporarily wet soil, contributing about half the total endosulfan remaining after 30 days.
Dissipation in surface 0-25mm soil was more rapid than in 0-50mm soil, but this varied from no
effect for the rapidly dissipated OP’s to 1.6 times faster for endosulfan, and varied according to Koc
for the herbicides. Herbicides with lower Koc dissipated faster in the surface layer than those with
higher Koc, due to greater downward movement. Downward movement decreased the apparent halflife
in the 0-25mm soil and increased the apparent half-life in 0-50mm soil. Half-lives in 0-25mm soil
were considerably lower than published ‘selected’ values. Dissipation of endosulfan was consistent
with studies in other warmer climates. The shallow depth of soil studied (which enhanced downward
movement) and application on the surface contributed to this more rapid dissipation. The results are
consistent with observations that “runoff available residues” dissipate more rapidly than generally
expected for bulk soil.
Endosulfan dissipated rapidly (e.g. 75-90%) from crop residue cover (wheat or cotton trash)
within the first day after spraying, apparently a result of volatilisation. Half-lives for endosulfan on
crop residues after the initial loss were similar to those in bare and covered soils. The data indicate
that a benefit of retaining crop residues on the soil surface, in addition to reducing runoff and sediment
losses, is that it intercepts and dissipates the endosulfan more rapidly than when sprayed onto soil.
Runoff extraction was investigated, in a simple empirical analysis, by comparing concentration
in soil (mg/kg) before rain and event average concentration in runoff (pg/L), using data from three
rainfall simulator studies in cotton fields, for 14 pesticides, and from the literature. The ratio of runoff
to soil concentrations, or the linear regression slope fitted through the origin, was termed the runoff
extraction ratio (ER0). The pesticides varied widely in solubility (0.003-700,OOOmg/L) and ranged
from strongly (DDE, KD~ 15,000) to weakly sorbed (fluometuron, dimethoate, pyrithiobac sodium, KD
<30). Runoff extraction behaviour from bare soil was remarkably consistent for pesticides of widely
different properties. Total concentrations in runoff of each pesticide were closely related to
concentrations in the soil (0-25mm) before rain, generally with a similar relationship for all pesticides
and sites, over four orders of magnitude range in concentrations. As a first approximation,
concentration in runoff (pg/L) = 28 times concentration in soil (mg/kg), (or Ero = 28). Runoff
extraction was also somewhat similar for dissolved N and P, and organic N.
Ero values were not related to partition coefficients (KP) measure in runoff. However, runoff
extraction did decrease with time after spraying and was lower for aged DDE and trifluralin at one
site. This is considered to relate to lower concentrations in the surface few mm of soil (c.f. 0-25mm
soil) over time. ERO values were similar for the slopes studied (0.2-4%), for long and short plots, and
for banded and blanket spray plots. Runoff extraction was reduced where cover reduces sediment
concentration. Runoff extraction was significantly lower for a weakly sorbed pesticide (dimethoate) in
only one instance and not for a range of other weakly sorbed pesticides at the other sites.
Concentrations in the water and sediment phases in runoff, and in sediment (mg/kg), were also
linearly related to soil concentrations for pesticides of similar KP, but extraction in two phases varied
according to normal partitioning (Eqn 5-5). The sediment concentration in runoff (10-60 g/L from
bare plots) had a secondary effect on ER0, and only affected ERO when sediment concentration was low
(i.e. with cover). This is because higher sediment concentrations were associated with lower
concentrations in the sediment (mg/kg), due to greater desorption and decreasing physical enrichment.
Less physical enrichment (due to size-selective sediment sorting) occurred than observed on
coarser textured soil (e.g. enrichment ratio up to 8), with enrichment ratios mostly less than 1.0 (due to
desorption) and no greater than 2.0. For all pesticides, the concentration in sediment (mg/kg) was
within a factor of about two of the soil concentration adjusted for desorption using the normal
partitioning equation. Organic carbon and clay were also only slightly enriched in sediment, despite
considerable deposition in the furrows. This is because the soils eroded as aggregates (due to low sand
and high clay content), and because coarser sediment had greater concentrations of sorbed pesticides
than finer sediment, the opposite of what is normally expected (e.g. where coarser sediment is sand).
The notable similarity of runoff extraction ratio for all pesticides in the rainfall simulator studies
was probably because (a) the main factor that limit runoff of weakly sorbed chemicals, i.e. leaching
from the runoff mixing zone, was ineffective because of low infiltration and ponding of infiltrated
water in the shallow tilled layer in the bottom of furrows, (b) sediment concentrations were high
enough to ensure transport of strongly sorbed pesticides, and (c) all pesticides had some transport in
both the water and sediment phases, diluting the response to sediment load. The concentration of
pesticide extracted from soil into runoff appears to be determined by the soil concentration, with, in
the absence of significant leaching and with sufficient sediment transport, little differentiation between
pesticides of different partition properties. This is partly because, on any plot, the same mass of soil and the same volume of water are involved in mixing, independent of the chemical being considered,
and because factors that increase extraction of solutes also tend to increase detachment of sediment.
Analysis of published runoff data for a range of pesticides in US croplands indicated similar
average runoff extraction to the rainfall simulator studies in Australian cotton fields. However, runoff
extraction was higher for much more erosive conditions (e.g. cultivated 10-15% slopes) and lower for
low erosion conditions (furrow irrigation on low sloping fields in California). Runoff extraction was
similar for this latter case (i.e. ERO~30) once adjusted to a higher sediment concentration.
Analysis of the rainfall simulator and published data presents a conceptual framework where the
major drivers of pesticide runoff were separated between (a) application rate and dissipation, described
by soil concentration at the start of rain, which accounts for five orders of magnitude differences in
runoff concentrations, and (b) runoff extraction during the rainfall event, which varied over a limited
range. The first of these factors causes most of the difference in runoff between pesticides.
Partition coefficients in runoff (KP) were not affected by cover and wheel traffic treatments even
though these treatments had large effects on pesticide runoff concentrations. KP values increased with
time after spraying, rapidly in the first few days and more slowly over the next few weeks, for all
pesticides. KP values were greater than soil sorption KD values, increasingly so for pesticides of lower
sorption. Thus pesticides normally considered weakly sorbed were much more sorbed in sediment
than expected, particularly at longer times. Conversely, moderately/strongly sorbed pesticides, such as
endosulfan, were less sorbed than expected in the first day or so.
Partitioning appeared to be influenced by both time of contact with soil and time of mixing
(during rain). The results are conceptually consistent with a two-compartment, bi-phasic (fast-slow)
sorption model, with the soil in the runoff-mixing layer under rainfall being a continuous dilution
system. The ‘slow’ phase, due to diffusion into less accessible soil domains, leads to increasing
partition coefficients with greater time of contact. The short time of mixing means that the water
phase is mainly interacting with the ‘fast’ or most accessible fraction, while the ‘slow’ fraction
remains in the sediment phase.
Percentages in the water phase in runoff, for 14 pesticides, roughly followed a published
relationship with solubility, and an empirical relationship with soil sorption Koc values, but only for
erosive conditions. These relationships do not reflect the full range in responses that occur due to the
likely range of concentrations and organic carbon content of sediment, or the increase in KP with time.
Because of lower KP values soon after spraying, less soluble pesticides had 20-45% in water.
Conversely, a few days/weeks after spraying, more soluble pesticides had only 60-80% in water. Thus
all pesticides tended to have a ‘foot in each camp’ and some potential for management using erosion
control practices.
It is an over simplification to expect ‘percent in water’ to be a characteristic of a pesticide. So
long as sediment concentration and KP can be estimated, the percentage in the water phase can be calculated quite simply from first principles (Eqn 5-5) and behaviour for relevant field conditions can
be assessed. This equation was used to show that reported values of percentage in the water phase for
endosulfan that appeared to conflict (20-95%) and the values from the rainfall simulator plots (15-
45%) are explained by differences in sediment concentration and organic carbon in the studies. A
wide range of percentages in water (10-95%) will occur for pesticides with KP of 5-500 (or solubilities
-1-100), such as endosulfan, for the range of sediment concentrations and organic carbon that might
occur in the environment.
Improved practices are needed to minimise soil erosion, and related agrochemical transport,
from cotton fields during rain. The most influential practice used in other agricultural industries, that
is, retaining crop residues as surface cover, is rarely practiced in the Australian cotton industry.
Therefore two options available to cotton growers, namely retention of surface cover and controlling
wheel traffic, were evaluated using simulated rain on a well-aggregated black Vertosol. Increasing
cover (0-60%) resulted in decreasing runoff, soil loss and sediment concentration. Runoff and soil
loss were reduced by an order of magnitude with about 50% cover and by a small amount with notraffic.
Cover and no traffic combined gave least runoff and soil loss.
Pesticide transport in runoff was also reduced strongly by retaining on-ground cover and
somewhat reduced by avoiding prior wheel traffic. With 45-60% cover, concentrations were reduced
5-fold for a-, |3- and total endosulfan; halved for endosulfan sulfate, trifluralin and DDE, and
unchanged for prometryn. Cover had more effect on endosulfan because cover intercepted and
dissipated the sprayed endosulfan, reducing concentrations in surface soil. Cover greatly reduced total
pesticide losses (g/ha) because cover reduced runoff and soil loss considerably. With 45-60% cover,
total losses were reduced by 90-98%. No-traffic gave 40% lower losses, and enhanced the effect of
cover, but did not prevent large pesticide losses from bare plots. Cover provided more control of more
soil-sorbed pesticides (endosulfan, trifluralin and DDE). Control of the less sorbed prometryn was
largely due to cover reducing runoff.
An examination of the practical requirements for maintaining effective cover in cotton farming
systems indicated that most of the perceived conflicts with insect, weed and irrigation management
could be overcome, although further study is needed.
Many of these results have only been possible because of the use of the rainfall simulator, multiresidue
pesticide analysis and the availability of sufficient resources. Such opportunities are rare in
field research. By allowing an intensive regime of runoff sampling at controlled times after pesticide
applications, the study has yielded data with more significance, enabling the conclusions made above
regarding the relative behaviour of individual pesticides and their extraction from soil in runoff. The
author acknowledges the contributions made by others to this study, but all of the experimental work
and the data reported in this thesis were under his control. | en_AU |
