We explore the ability of weak lensing surveys to locate massive clusters .
We use both analytic models of dark matter halos and mock weak lensing surveys generated from a large cosmological N -body simulation .
The analytic models describe average properties of weak lensing halos and predict the number counts , enabling us to compute an effective survey selection function .
We argue that the detectability of massive halos depends not only on the halo mass but also strongly on redshift at which the halo is located .
We test the model prediction for the peak number counts in weak lensing mass maps against the mock numerical data , and find that the noise due to intrinsic galaxy ellipticities causes a systematic effect which increases the peak counts .
We develop a correction scheme for the systematic effect in an empirical manner , and show that , after the correction , the model prediction agrees well with the mock data .
The mock data is also used to examine the completeness and efficiency of the weak lensing halo search with fully taking into account the noise and the projection effect by large-scale structures .
We show that the detection threshold of { S / N } = 4 \sim 5 gives an optimal balance between completeness and efficiency .
Our results suggest that , for a weak lensing survey with a galaxy number density of n _ { g } = 30 arcmin ^ { -2 } with a mean redshift z = 1 , the mean number of halos which are expected to cause lensing signals above { S / N } = 4 is N _ { halo } ( { S / N } > 4 ) = 37 per 10 deg ^ { 2 } , whereas 23 of the halos are actually detected with { S / N } > 4 , giving the effective completeness as good as 63 % .
On the other hand , the mean number of peaks in the same area is N _ { peak } = 62 for a detection threshold { S / N } = 4 .
Among the 62 peaks , 23 are due to halos with the expected peak height { S / N } > 4 , 13 are due to halos with 3 < { S / N } < 4 and the remaining 26 peaks are either the false peaks due to the noise or halos with a lower expected peak height .
Therefore the contamination rate is 42 % ( this could be an overestimation ) .
Weak lensing surveys thus provide a reasonably efficient way to search for massive clusters .