The authors present an analytical climate model, which has the feature
s that (i) the atmosphere is a simple oscillator for all periods great
er than or equal to I year, (ii) the ocean stores heat, (iii) the ocea
n exchanges momentum with the atmosphere, and (iv) random forcing exis
ts due to atmospheric thermodynamics and oceanic dynamics. The piecewi
se analytical integration of coupled linear equations for sea temperat
ure, air-sea temperature difference, and air-sea velocity difference g
enerates experimental climates. The exchange parameters of the algorit
hm. except for the exchange coefficient for heat with the deep ocean,
are calibrated to the observed climate using the annual cycle, and ran
dom forcing is applied over intervals of one year, The atmospheric ran
dom forcing leads to bounded random walks, the extent of which increas
es as the exchange coefficient with the deep ocean decreases, and the
oceanic random forcing generates a stationary response. It is found th
at the observed statistics of the global temperature series can be rep
roduced by either a relatively large heat exchange coefficient with th
e deep ocean and little oceanic variability or a smaller exchange coef
ficient with a larger oceanic variability. Plausible exchange coeffici
ent values imply random walk lengths of at least a century-long timesc
ale.