Wat. Res. Vol. 35, No. 13, pp. 3168–3178, 2001 # 2001 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0043-1354/01/$-see front matter PII: S0043-1354(01)00011-2 IN SITU MEASUREMENT AND STATISTICAL MODELLING OF ESCHERICHIA COLI DECAY IN SMALL RIVERS PASCAL BEAUDEAU 1 *, NICOLAS TOUSSET 1 , FRANCK BRUCHON 2 , AME ´ LIE LEFE ` VRE 1 and HUW D. TAYLOR 3 1 Laboratoire d’Etudes et d’Analyses de la Ville du Havre 70 quai Frissard, 76600 Le Havre, France; 2 Agence de l’Eau Seine-Normandie-Direction des Rivages Normands 21 rue de l’Homme de Bois, 14600 Honfleur, France and 3 The School of the Environment, Brighton University Cockcroft Building, Lewes Road, Brighton BN2 4GJ, UK (First received 17 July 2000; accepted in revised form 18 December 2000) Abstract}Numerous studies have been carried out on the environmental factors associated with the decay of faecal bacteria in open (fresh or marine) waters. The present study aimed at understanding the fate of bacteria in small streams (flow 520m 3 s 1 ) for which there is a lack of knowledge. An original in situ protocol was developed for measuring the die-off of Escherichia coli (E. coli) from wastewater treatment plants. Based upon 80 values of the decay first-order parameter (K or its inverse T90), collected from five rivers in Normandy (France), a median T90 of 10h and a minimal T90 of 1.3h were obtained. K was then modelled as a linear function of variables made up from flow, water temperature and suspended particulate matter (SPM). The set of significant co-variables did not include light indicators. E. coli decay is inversely related to the river flow and it becomes highly significant below 0.3m 3 s 1 . The positive effect ofsmallflowsondie-offisincreasedbywatertemperatureover158C,whereasitcouldbereducedbySPM. The major co-variable of the model ( p510 9 ) is an empiric composite variable integrating the effect of flow and temperature that explains more than 40% of the variance of K. We interpreted this as an expression of predation by benthic micro-grazers which could be the main cause of E. coli die-off in small streams in temperate countries. # 2001 Elsevier Science Ltd. All rights reserved Key words}Escherichia coli, rivers, decay model, flow, temperature, SPM INTRODUCTION Microbial contamination from faecal origin in natural waters causes a sanitary risk through consumption of drinking water, consumption of shellfish and bathing. However, as it disappears spontaneously, it is important to quantify its kinetics of decay, for assessing the level of sewage treatment required for the protection of sensitive areas. Our study focused on bacterial indicators, since European sanitary regulations are based on the monitoring of this kind of indicator. Escherichia coli (E. coli) was chosen because of its broad use as a tracer for faecal contamination. Usually, population dynamics for allochtone bacteria are modelled by first-order kinetics: dNðtÞ=dt ¼kNðtÞ with k > 0, i.e. an exponential decay of N with time t. This model is both simple and efficient in some practical situations. Its single parameter, i.e. the decay coefficient k, is often replaced by K ¼ k=2:3, which corresponds to the use of decimal logarithms for bacterial counts: log NðtÞ log Nðt 0 Þ¼K ðt  t 0 Þ. K , usually ex- pressed in h 1 , and is thus the inverse of the period of time T 90ðhÞ necessary for reducing the bacterial population by 90%. The decay coefficient varies with the nature bacteria and their physiological state, but also with environmental factors. Numerous studies have fo- cused on the environmental parameters influencing K ,bothincoastalwatersandinsomelargerivers(see Barcina et al., 1997 for a review), but no specific work has been done on small streams. This study aimed at evaluating the distribution of K in small streams (flow within the range 0.1–20m 3 s 1 ), and at pointing out the main environmental decay factors in this type of environment. Use of laboratory experi- mental set-ups or microcosms set up in sea or river locations was not acceptable here, as the goal was to study the combined expression of hydrological and ecological factors specific to small streams. This experimental in situ approach required the develop- ment of original tools for measuring the decay coefficient. This paper details both the field experiment used for data collection and the statistical model *Author to whom all correspondence should be addressed. Tel.:+33-141-796822; fax:+33-141-796768; e-mail: p.beaudeau@invs.sante.fr 3168