Field evaluation of passive samplers: Monitoring polycyclic aromatic hydrocarbons (PAHs) in stormwater

The aim of this study was to evaluate the feasibility of using semipermeable membrane devices (SPMDs) and polyethylene-based passive sampler devices (PSDs) for monitoring PAHs in stormwater. Firstly, SPMDs were deployed at one site and SPMD-derived water concentrations were compared with water concentration measured from grab samples. In a subsequent deployment the performance of SPMDs and PSDs was compared. Finally PSDs of multiple surface area to volume ratios were used to compare PAH concentrations at the two sites. The results obtained in this study show that SPMDs can be used to measure the water concentration of PAHs with reasonable accuracy, when compared with grab samples collected at the same site. Importantly, several PAHs which could not be detected in a 10 L grab sample could be detected in the SPMDs. PSD and SPMD samplers produced similar results when deployed at the same site, with most estimated water concentrations within a factor of 1.5. The use of PSDs in multiple surface area to volume ratios proved to be an effective means of characterizing the uptake kinetics for PAHs in situ. Overall passive water samplers proved to be an efficient technique for monitoring PAHs in stormwater.

Handling varying amounts of missing data when classifying mental-health risk levels

One of the main challenges of classifying clinical data is determining how to handle missing features. Most research favours imputing of missing values or neglecting records that include missing data, both of which can degrade accuracy when missing values exceed a certain level. In this research we propose a methodology to handle data sets with a large percentage of missing values and with high variability in which particular data are missing. Feature selection is effected by picking variables sequentially in order of maximum correlation with the dependent variable and minimum correlation with variables already selected. Classification models are generated individually for each test case based on its particular feature set and the matching data values available in the training population. The method was applied to real patients' anonymous mental-health data where the task was to predict the suicide risk judgement clinicians would give for each patient's data, with eleven possible outcome classes: zero to ten, representing no risk to maximum risk. The results compare favourably with alternative methods and have the advantage of ensuring explanations of risk are based only on the data given, not imputed data. This is important for clinical decision support systems using human expertise for modelling and explaining predictions.