In SimBiology, compartments can also represent 2-dimensional areas, such as the surface of a cell. I believe the error you are seeing is not really about any 2-dimensional compartments, but it part of a larger error message indicating that you need to write your reaction rate with units of amount per time. The full error message should look something like this:
"The reaction rate for reaction ... must be defined with dimensions of amount per time not concentration per time. SimBiology converts each reaction rate to dimensions of amount per time, even when dimensional analysis is disabled. SimBiology cannot identify a compartment volume for this conversion when a reaction has reactants in multiple compartments, or when the reactant is null and the reaction has products in multiple compartments."
Usually, this means you need to multiply your rate constants by the appropriate volumes or areas. Here's a simple example based on the model in this paper. Consider a reaction for a surface receptor binding with a ligand to form a surface complex: Surface.Receptor + Volume.Ligand <-> Surface.Complex. Regardless of the units used to measure Receptor, Ligand, and Complex, you will need to write the reaction in units of amount/time, in this case molecule/minute. In the paper, Receptor and Complex are measured in molecule, and Ligand is measure in nanomole. The forward reaction in units of molecule/time is k_on*Receptor*Ligand, where k_on has units of 1/nanomolar/minute. The reverse reaction in units of molecule/time is k_off*Complex, where k_off has units of 1/minute. So in SimBiology, you would write this reaction rate as k_on*Surface.Receptor*Volume.Ligand - k_off*Surface.Complex. SimBiology then knows how to convert this to the appropriate rate of change for Ligand (specifically, it knows to divide by Avogadro's number and the cellular volume).
I hope this example clears things up. If not, I recommend making a new MATLAB Answers question that provides the exact reactions and rates that you are having trouble converting. It would also be helpful if you can attach a file that allows us to see how you've currently implemented the model.