A theoretical analysis of the angular resolution of two-dimensional in
terferometric radiometers for Earth observation from low-orbit satelli
tes and its degradation due to spatial decorrelation effects is presen
ted. The analysis extends basic effects known in the context of radio
astronomy (application with narrow field of view, very few baselines)
and in one dimension (ESTAR L band, few baselines interferometric radi
ometer) to the wide-field-of-view, many-baseline, high-resolution two-
dimensional system required by Earth observation applications and comp
utes beam width, encircled energy (or main beam efficiency), and side
lobe level as a function of windowing (apodization) to allow for an op
timum angular versus radiometric resolution trade-off. It is found tha
t the extension of the Barlett window (which has a poor performance in
one dimensional signal processing) to two dimensions produces high-qu
ality results, comparable or better than those of Gaussian and Blackma
nn windows. Theory is extended to hexagonally sampled systems based on
a Delta or Y-shaped instrument, with hexagonal- and star-shaped suppo
rt regions in the visibility space, respectively. The superior perform
ance of the latter over the former for the same number of antennas and
correlators is quantified and details of the angular resolution of on
e instrument of this kind, MIRAS, under development by the European Sp
ace Agency, are presented. For this radiometer Gaussian or Barlett win
dows should be used for good radiometric sensitivity or spatial resolu
tion, respectively. In both cases the effects of decorrelation within
the small alias-free field of view are negligible. It is also found th
at the impact of hardware imperfections which exist within the strict
requirements of the specifications have a negligible effect on the ang
ular resolution. Finally, experimental angular resolution results with
a laboratory breadboard in a focused near-field setup are presented a
nd compared to the theoretical predictions.