A spatial, statistical approach to map the risk of milk contamination by β-hexachlorocyclohexane in dairy farms

• Main Authors: Sabrina Battisti, Antonino Caminiti, Giancarlo Ciotoli, Valentina Panetta, Pasquale Rombolà, Marcello Sala, Alessandro Ubaldi, Paola Scaramozzino
• Format: Article
• Language: English
• Published: PAGEPress Publications 2013-11-01
• Series: Geospatial Health
• Subjects: beta-hexachlorocyclohexane, geostatistical analysis, indicator kriging, bulk milk, Italy.
• Online Access: http://www.geospatialhealth.net/index.php/gh/article/view/56

In May 2005, beta-hexachlorocyclohexane (β-HCH) was found in a sample of bovine bulk milk from a farm in the Sacco River valley (Latium region, central Italy). The primary source of contamination was suspected to be industrial discharge into the environment with the Sacco River as the main mean of dispersion. Since then, a surveillance programme on bulk milk of the local farms was carried out by the veterinary services. In order to estimate the spatial probability of β- HCH contamination of milk produced in the Sacco River valley and draw probability maps of contamination, probability maps of β-HCH values in milk were estimated by indicator kriging (IK), a geo-statistical estimator, and traditional logistic regression (LR) combined with a geographical information systems approach. The former technique produces a spatial view of probabilities above a specific threshold at non-sampled locations on the basis of observed values in the area, while LR gives the probabilities in specific locations on the basis of certain environmental predictors, namely the distance from the river, the distance from the pollution site, the elevation above the river level and the intrinsic vulnerability of hydro-geological formations. Based on the β-HCH data from 2005 in the Sacco River valley, the two techniques resulted in similar maps of high risk of milk contamination. However, unlike the IK method, the LR model was capable of estimating coefficients that could be used in case of future pollution episodes. The approach presented produces probability maps and define highrisk areas already in the early stages of an emergency before sampling operations have been carried out.