We present a detailed analysis of the star formation history ( SFH ) of three fields in M33 located \sim 4 - 6 visual scale lengths from its nucleus .
These fields were imaged with the Advanced Camera for Surveys on the Hubble Space Telescope and reach \sim 2.5 magnitudes below the red clump of core helium burning stars .
The observed color-magnitude diagrams are modeled as linear combinations of individual synthetic populations with different ages and metallicities .
To gain a better understanding of the systematic errors we have conducted the analysis with two different sets of stellar evolutionary tracks which we designate as Padova ( Girardi et al .
2000 ) and Teramo ( Pietrinferni et al .
2004 ) .
The precise details of the results depend on which tracks are used but we can make several conclusions that are fairly robust despite the differences .
Both sets of tracks predict the mean age to increase and the mean metallicity to decrease with radius .
Allowing age and metallicity to be free parameters and assuming star formation began \sim 14 Gyr ago , we find that the mean age of all stars and stellar remnants increases from \sim 6 Gyr to \sim 8 Gyr and the mean global metallicity decreases from \sim - 0.7 to \sim - 0.9 .
The fraction of stars formed by 4.5 Gyr ago increases from \sim 65 \% to \sim 80 \% .
The mean star formation rate 80 - 800 Myr ago decreases from \sim 30 \% of the lifetime average to just \sim 5 \% .
The random errors on these estimates are \sim 10 \% , 1.0 Gyr , and 0.1 dex .
By comparing the results of the two sets of stellar tracks for the real data and for test populations with known SFH we have estimated the systematic errors to be 15 \% , 1.0 Gyr , and 0.2 dex .
These do not include uncertainties in the bolometric corrections or variations in \alpha -element abundance which deserve future study .