The spatial autoregressive process X_k,ℓ = (X_k-1,ℓ + X_k,ℓ-1) + ε_k,ℓ, where k, ℓ ≥ 1 is investigated. We consider the least squares estimator ˆ_m,n of based on the observations X_k,ℓ: 1 ≤ k ≤ m and 1 ≤ ℓ ≤ n. In the stable (i.e. asymptotically stationary) case, when || < 1/2, asymptotic normality   as m, n → ∞ with m/n → constant > 0 can be derived from the previous more general results due to Basu and Reinsel (1992, 1993, 1994). In the unstable case, when || = 1/2, we prove again asymptotic normality, but (in contrast to the doubly geometric spatial model) with a surprising rate of convergence, namely   as m, n → ∞ with m/n → constant > 0.