This project focuses on fish population dynamics modeling and statistical inference. Survey data for Atlantic Herring, Cod, Silver Hake, Spiny Dogfish, and Yellowtail Flounder are used for analysis, which were collected twice a year from 1968 to 2013 by NOAA, once in the spring and once in the fall, and were labeled as south or north based on the location it was collected. First, state-space model with AR(1) process, state-space model with seasonality and trend, and SARIMA models are applied to Cod and Herring data set, and state-space models perform better in most cases based on measurements on fitting and prediction performance. To improve modeling, threshold model is combined with state-space model with AR (1) process. From results of applying state-space models, we find that if the fish abundance shows a continuous increasing or decreasing trend, we expect that the state-space with seasonality and trend model will perform better. And if the abundance shows a changed trend over time, we expect that state-space with threshold and AR(1) model will perform better. In our models, we only take data patterns into consideration without any other features such as temperature and fishing regulations, which could be combined with in the future and make a better fitting and prediction.