This paper sought  to  compare  the  performance of different  distribution models in  flood forecasting  using  twenty  six years  rainfall  data  of  the  study  area  (Zaria)  as  input variables. The study is  necessitated following  past flooding  events in  the  area.Gumbel, Log  Pearson  T ype  III  and  Chegodayve's  distribution  models  were  employed  in  the modeling with the  view  of  recommending  the  best  fitting  curve  for  the  area.  Extreme values of rainfall data obtained fr om Nigerian Meteorological  Agency,  Aviation Zaria, were used for the analyses.From the result of analyses of the twenty six years rainfall data  (1989-2014),  using  different  distribution  models to  predict  rainfall  depth  that  may cause flood  in  the  area when compared to  the  true  meteorological readings of the  area, Log – Pearson T ype III  model produced the  greatest correlation coefficient (0.90)  as well as least deviation (0.1507). The average annual rainfall (AAR) for the twenty six years return periods for Gumbel, Chegodayve and Log Pearson are 1553.59, 1389.39 and 1 161.69mm respectively .  Based on the AAR values,Log Pearson's produced AAR that  is  nearer  the  meteorological  value  of  1034.34mm.  The error  difference  for  Gumbel and  Chegodayve  are  15%  and  20%  respectively  in  terms  of  their  correlations  with respect  to  Log  Pearson's.At  any  return  period  (X),  based  on  the  model,  the  rainfall  depth can be determined and compared with the  available meteorological values, for  flood prediction  and  for ecasting  in  the  area.  It  is  recommended  that  more  gauging  stations  be installedin  Zaria  so  as  to  have  a  wider  coverage  and  a  model  that  will  simulate  the  entire
catchment.

KEYWORDS:  Probability  distribution  model,  Recurrence  interval,  Average  annual
rainfall,  Gumbel,  Log  Pearson,  Chegodayve,  Meteorological  readings.