This paper presents a procedure to model texture fields using seasonal autoregressive, moving average models. The modeling of 2-D images has been formulated as a 1-D time series analysis problem. Properties such as directionality and clustering have been fully investigated and presented. The applications of this 1-D seasonal ARMA process to texture analysis, synthesis and data compression have been discussed. It was demonstrated that a cloud field image can be quantitatively defined and its surrogates can be synthesized by the model parameters. The implications for the quantitative study of cloud climatology is thus evident.