Computational physics

A simulation divides a domain into regions and samples separately within each region. The regions have different areas. How should their estimates generally be combined to estimate a domain-wide average?