) used to support pesticide management decisions at the catchment-scale can deliver environmental protection while retaining farm production benefits. Presently, ERA’s in Australia are typically performed to evaluate farm-scale ecological impacts from pesticides. However, as pesticide exposure in rivers is usually a result of the activities of more than one farmer affected by spatially and temporally explicit factors such as climate, hydrology, geomorphology and land uses, a catchment-scale pesticide management approach seems a logical progression. The aim of this thesis was to investigate the potential effectiveness of applying ERA adapted to the catchment-scale as a pesticide management tool for agricultural catchments.

Initially, a catchment-based ERA of diuron, prometryn and endosulfan use in the Gwydir River catchment, NSW, Australia, was used to identify possible aquatic exposure sites. The classic phases of problem formulation, analysis and risk characterisation were established. The problem formulation phase identified hazard concerns and established assessment endpoints specific for areas that were considered to have high (95% of species protected 95% of the time) or lower (90% of species protected 95% of the time) ecological value. The analysis phase identified that the likely exposure sources of diuron, prometryn and endosulfan in the reaches of the Gwydir River catchment were agriculture, characterised using available ecotoxicology and regulatory exposure monitoring information using continuous probability distributions.

The characterisation of risk involved comparing distributions of exposure and ecotoxicity as species sensitivity distributions (SSDs) to estimate the probabilities that the endpoints were being exceeded (Solomon et al., 2000). With the exception of prometryn, significant risk from diuron (maximum risk = 8.95%) and endosulfan (maximum risk = 7.86%) exposure was found to occur in some reaches of the Gwydir River catchment. These areas, considered as “hot spots”, predominated where intensive agricultural production was most prevalent.

An uncertainty evaluation identified a number of information shortfalls in this ERA. These gaps included permanency of ecological effect resulting from pulse exposures, conservative risk estimation that was based on exposure data from a sampling regime targeted when chemical use and rainfall were more prevalent, and, for the purpose of supporting risk management, identification of specific areas in the catchment contributing chemical loads in areas of concern. These uncertainties led to the need for further research.

The consideration of organism recovery under a pulse exposure scenario likely to be observed in the Gwydir River catchment was investigated in a laboratory toxicity study. This study tested the potential for two duckweed species (Lemna minor and L. gibba) to recover from a seven day diuron pulse. The duckweed species were exposed to a range of diuron concentrations (0.3-200 μg L-1) for seven days, and placed in to clean growth media for a further seven days to simulate a recovery phase.

Exposure toxicity and recovery were evaluated through plant and frond counts, and wet and dry weights. The inhibition of growth was used as the toxicity metric by comparing growth response with control (0 μg L-1) populations. Significant growth inhibition of L. minor (EC50 = 34.9 and 52.8 μg L-1 for dry weight and frond count, respectively) and L. gibba (EC50 = 50.3 and 47.6 μg L-1 dry weight and frond count, respectively) was observed at the end of the seven day exposure. By the end of the seven day recovery phase, growth inhibition compared to the controls were shown to decline for a range of treatment exposure concentrations to the point that inhibition was not significantly different from the control for both L. minor (50 and 100 μg L-1, for dry weightand frond count, respectively) and L. gibba (200 and 50 μg L-1, for dry weight and frond count, respectively). With reference to the literature, growth inhibition was determined to be in response to photosynthesis inhibition (Haynes et al., 2000; Fai et al., 2007). Population recovery for treatment concentrations greater that observed in the Gwydir River catchment suggested was a clear reversal of this effect. It was concluded that both L. minor and L. gibba could sufficiently recover from a prolonged exposure event, suggesting the possibility of keystone aquatic plants and algae resilience to diuron exposure occurring in the Gwydir River catchment.

To clarify the risk characterisation uncertainty of diuron in the Gwydir River catchment, a spatial exposure modelling framework was required. A modelling procedure with the capacity to provide a daily time series exposure concentration pointing to catchment areas contributing to chemical load was selected. This framework involved combining two models, a chemical fate model (Pesticide Root Zone Model, PRZM) and a chemical routing model (Riverine Water Quality model, RIVWQ). The inputs to these models were obtained and processed from readily accessible databases and/or literature. Required inputs included soil, land use and weather station information; characteristic agronomic practices for different land uses and their respective label application rates. In accordance with the chemical labels all maximum application rates were used for the respective land uses of cotton, wheat, chickpea, canola and pasture in the simulations. To account for the full range of in season applications two scenarios were required to be run. Specifically, pre- (i.e. chemical incorporated in the top 4 cm of soil) and post-emergence (chemical applied directly to the surface) were simulated for cotton.

The simulation results showed that under the post-emergence application regime the highest chemical loading for streams was predicted. This was the result of chemical being more readily available at the surface to be entrained in runoff. It was found that the post-emergence application scenario reflected more closely the peak concentration magnitude and timing observed in the monitoring data. However, when compared with monitoring data, the model framework was unable to predict peak exposure concentrations on precisely the same dates. This indicated a degree of error in the model predictions, an outcome likely to be the result of uncertainties in the model inputs, especially with respect to pesticide applications, timing and crop rotation scenarios. However, the modelling framework did perform in a way that was consistent with chemical fate and fugacity principles. Subsequently, from these different scenarios, the sub-catchments contributing chemical loads were able to be identified, potential pulse durations were characterised as were their probabilities of re-occurrence, with the longest pulse exceeding the toxicity threshold lasting 6-9 days. It was concluded that this approach to estimating exposure at the sub-catchment level could be a useful tool for catchment management, devising and directing risk management strategies associated with monitoring. However, this will require further calibration and validation for effective use as a risk management tool.

The outputs of this thesis are suggested to provide justification for further development of catchment-based ERA strategies in Australia. These would include site specific chemical loading,employ probabilistic risk characterisation and account for ecosystem biodiversity value and resilience. This strategy would take ERA in Australia from the top-down approach now taken by national regulators to a bottom-up alternative, inclusive of catchment managers

operating at the local level interacting directly with stakeholders. Pesticide exposure concerns identified through an ERA should then be addressed through the implementation of a management strategy. This strategy would utilise outputs from spatial modelling supported by monitoring, as a basis for directing management to areas of a catchment where it is most needed. The thesis concludes that managing pesticide exposure in agricultural catchments with more informed ERA can provide a sounder basis for pesticide use in crop production coexisting with ecosystem protection.