There is a potential risk that hazardous radioactive sources could enter the environment, e.g., via satellite debris, smuggled radioactive goods, or lost metal scrap. From a radiation protection point of view there is a need for rapid and reliable methods for locating and identifying sources. The methods could also be used to locate hot spots after radioactive fallout. Carborne and airborne gamma spectrometry systems are suitable for the task. This work focuses on a situation where the radionuclide to search for is known, which is not an unlikely scenario. The possibility that the source is located near a road can be high, and thus motivating a carborne spectrometer system. The main object is to optimize on-line statistical methods in order to achieve a high probability for locating the point source and still have reasonably few false alarms caused by variations in the natural background radiation. Data were obtained from a carborne 3-L NaI(Tl) detector and two point sources located at various distances from the road. The nuclides used were 137Cs and 131I. Spectra were measured stationary on the road. From these measurements we have reconstructed counts in spectral windows applicable to different speed and sampling times; the time 3 s and speeds 32 and 54 km h(-1) are used in this work. The maximum distance a source can be located from the road and still be detected is estimated with four different statistical analysis methods. This distance is called the critical distance, CD. The method is applied on gross counts in the full energy peak spectral window. For each method alarm levels have been calculated from background data obtained in Scania (Skåne), in the south of Sweden. The results show large differences in CD. With the best approach, the two sources could be detected from about 180 m (137Cs, 6 GBq) and 170 m (131I, 4.5 GBq).