The one-dimensional treatment of turbulent convection had large successes until the early 2000s. However, the recent abundance and precision of observational data shows that this problem is far from solved. Even so, ongoing theoretical debates about proper one-equation-based treatment of convection and new results show that it has various other theoretical difficulties as well. A more modern approach should be developed by using multidimensional models. We established a new theoretical framework for comparison between one-dimensional and multidimensional convection models by mapping the two-dimensional structure of the convective zone and optimizing the modeling parameters of the SPHERLS code. We constructed a series of static envelope models for the same RR Lyrae stars, but with different horizontal sizes and resolutions. We then used a series of statistical methods to quantify the sizes of convective eddies, map the energy cascade, and describe the different structural parts of the convective zone. These include integral length scales, Fourier series, and the determination of the convective flux through horizontal averaging. The structure of the convective zone depends significantly on the model size below an angular size of . Models of at least this size are more consistent, and the horizontal resolution of earlier studies is adequate to describe the granulation pattern in the large eddy simulation approach. In quasi-static RR Lyrae stars, the convective zone consists of two distinct dynamically unstable regions that are loosely connected. Approximately half of the convective flux is supplied by the transport of ionization energy in the partial hydrogen ionization zone. The 2D models presented in this work with the described size and resolution parameters can be used for comparison against 1D models. The structure of the convective zone urges reconsideration of some recent approaches to describe the convective flux currently used in radial stellar pulsation codes, which will be addressed in a separate paper.