Dear Sir, Yes, you can use Gumble Method for Discharge Measurement as well as Flood frequency analysis, but at least 10 ya 15-year data required for analysis.
If you want to perform flood frequency analysis (by processing annual maximum flow series) using the gumbel distribution/method, then the data length may not be adequate. But you can still perform analysis using 10 data points. Only problem would be regarding the uncertainty in the parameter estimates of the distribution. The uncertainty in the parameter estimation would be more if less data is available. The problem of having inadequate data for frequency analysis can be solved by using either of the alternantives:
1] You can use methods like profile likelihood function or Bayesian approach to account for less amount of data. These methods would try to provide uncertainty associated with the discharge estimates (i.e. quantiles from various return periods).
OR
2] You can use regional frequency analysis to pool data from hydrologically similar sites/gauges and then perform frequency analysis using pooled data.