Where Level of Detail (LoD) allow City Geography Markup Language (CityGML) models to be shared over different application domains, only one LoD defines building interiors. As there are potentially numerous applications with a need for building interiors (but not with all the details as defined in LoD4), in order to attain a larger applicability the LoD system should be extended with building interiors in accordance with the existing exterior definitions. Since CityGML LoD2 is a common LoD, in this research it is investigated how to automatically generate interiors for LoD2 (resulting in a model called LoD2+) in which the details are comparable to those of the exteriors. A key application in the Netherlands for LoD2+ is the determination of net internal area of buildings (the surface area which can be used by building owners). This property is useful for many applications which require the size of buildings, but is especially important for real estate taxations and sales. The registration of these values by Dutch municipalities (via Basisregistratie Adressen en Gebouwen (BAG) - Key Register for Addresses and Buildings) is at the moment still a manual process (determined from building blueprints), and is therefore error-prone. This research dealt with the design of a finite set of shape rules to generate valid LoD2+ city models from a CityGML LoD2 model with a small amount of semantic information. Here LoD2+ is defined as the existing building solids complemented by inner shells for each storey in a premises where space is allocated for walls, roofs, ceilings and floors. Using Boolean set operations (e.g. union, intersection and difference) the building exteriors can be split in the right amount of storeys according to BAG and erosion can be applied to offset walls, roofs, ceilings and floors to obtain storey solids according to LoD2+. From LoD2+ buildings in which interior surfaces are classified, the net internal area of the majority of premises can be estimated. By calculating these values from LoD2+ city models a comparison can be done, resulting in the validation of both the net internal area in BAG and the LoD2+ city model. Significant errors are found in the BAG registration, since net internal area was mostly not recalculated after the net internal area standards were changed/introduced. For this research prototype software has been developed in C++ for the conversion of LoD2 to LoD2+ buildings and the calculation of net internal area. This software package is open source and freely available at http://lod2plus.googlecode.com.