DETERMINISTIC VERSUS STOCHASTIC QUANTITATIVE MICROBIAL RISK ASSESSMENT MODELS FOR WASTEWATER IRRIGATION OF FOOD CROPS

Abstract

Listen

Translate article

Reuse of wastewater to irrigate food crops is being practiced in many parts of the world and is becoming more commonplace as the competition for, and stresses on, freshwater resources intensify. But there are risks associated with wastewater irrigation, including the possibility of transmission of pathogens causing infectious disease, to both workers in the field and to consumers buying and eating produce irrigated with wastewater. To manage these risks appropriately we need objective and quantitative estimates of them. This is typically achieved through one of two modelling approaches: deterministic or stochastic. Each parameter in a deterministic model is represented by a single value, whereas in stochastic models probability functions are used. Stochastic models are theoretically superior because they account for variability and uncertainty, but they are computationally demanding and not readily accessible to water resource and public health managers. We constructed models to estimate risk of enteric virus infection arising from the consumption of wastewater-irrigated horticultural crops (broccoli, cucumber and lettuce), and compared the resultant levels of risk between the deterministic and stochastic approaches. Several scenarios were tested for each crop, accounting for different concentrations of enteric viruses and different lengths of environmental exposure (i.e. the time between the last irrigation event and harvest, when the viruses are liable to decay or inactivation). In most situations modelled the two approaches yielded similar estimates of risk (within 1 order-of-magnitude). The two methods diverged most markedly, up to around 2 orders-of-magnitude, when there was large uncertainty associated with the estimate of virus concentration and the exposure period was short (1 day). Therefore, in some circumstances deterministic modelling may offer water resource managers a pragmatic alternative to stochastic modelling, but its usefulness as a surrogate will depend upon the level of uncertainty in the model parameters.