The daytime variation of hydroperoxy (HO2) radical concentration was observed by an instrument based on laser-induced fluorescence with NO addition at Oki Island, Japan, in July/August 1998. Although OH was not detected due to the high detection limit of the instrument, HO2 was determined with the detection limit of 0.8 parts per trillion by volume (pptv) (S/N=2, integration time of 1 min). On average, HO2 showed a maximum concentration of around 9 pptv in the early afternoon hours. During the field campaign, chemical species and meteorological parameters such as O3, CO, nonmethane hydrocarbons (C2-C6), NO/NO2, HCHO/CH3CHO, SO2, HNO3 and J(NO2) were also observed. Unfortunately, the absolute value of J(O1D) was not measured. Model calculations for radical concentrations were performed and they were compared to the observed hourly HO2 concentrations. On August 9 the calculated HO2 matched the observation very well within the uncertainties of observations (¿26%, 1&sgr;) and the model (¿24%, 1&sgr;). This indicates a good performance of model calculations in estimating HO2 under certain conditions with plenty of isoprene. On other days, however, model usually overestimated HO2 by a factor of 2, especially in the hours around noon. It is deduced that some important HOx loss chemistry is missing in the model. Although the cause of the discrepancy is not fully understood, possible mechanisms to explain the overestimation are studied. A hypothesis with additional loss of HO2 would be more plausible than that with additional OH loss. The additional loss rate for HO2 that can reduce the calculated HO2 to the measured level was computed for each hour. Its variation correlated well with those of HO2 and some photochemical products. The possibilities of HO2 loss on aerosol surface, unknown acceleration of HO2 reactions in the presence of high H2O, and HO2 reactions with carbonyl species are discussed. ¿ 2000 American Geophysical Union