A new multiple source detection threshold criterion based on peak-to-average power ratio(PAR)is proposed to solve the problem that multiple source detection methods based on information criterion cannot correctly estimate the number of sources under low signal-to-noise ratio in underwater environment(PARTC). The underwater array receiving data are weighted using the eigenvectors of the data correlation matrix

and the weighted data are calculated. Then

PARs of the weighted data are computed

and then sorted in a descend order. The number of sources can be estimated using the information that the PARs of pure noise have a linear distribution

and that the average grad of PARs is independent of snapshots

and when there exists signals the grad of signal PARs will be larger than the average grad of pure noise PARs. Simulation results on an 8-elements array show that the probability of correct detection of PARTC is larger than those of both the Akaike information criterion method and the minimum description length method under the conditions of low signal-to-noise ratio

small snapshots and closely spaced sources

while the calculation of PARTC is not expensive. Thus PARTC has board applications in underwater environment.