We present the first analysis of global and hemispheric surface warming tre
nds that attempts to quantify the major sources of uncertainty. We calculat
e global and hemispheric annual temperature anomalies by combining land sur
face air temperature and sea surface temperature (SST) through an optimal a
veraging technique. The technique allows estimation of uncertainties in the
annual anomalies resulting from data gaps and random errors. We add indepe
ndent uncertainties due to urbanisation, changing land-based observing prac
tices and SST bias corrections. We test the accuracy of the SST bias correc
tions, which represent the largest source of uncertainty in the data, throu
gh a suite of climate model simulations. These indicate that the correction
s are likely to be fairly accurate on an annual average and on large space
scales. Allowing for serial correlation and annual uncertainties, the best
linear fit to annual global surface temperature gives an increase of 0.61 /- 0.16 degreesC between 1861 and 2000.