Every simulation system has limitations due to the choices made in the design and development of the software. The limitations that arise in this program are due to two aspects of the design: Simplified Model Formulation, and Simplified Flow Representation. Simplifying the model formulation allows the program to complete simulations very quickly while producing accurate and precise results. Simplifying the flow representation aids in keeping the compute process efficient and reduces duplication of capability in the HEC software suite.

## Model Formulation

All of the mathematical models included in the program are deterministic. This means that the boundary conditions, initial conditions, and parameters of the models are assumed to be exactly known. This guarantees that every time a simulation is computed it will yield exactly the same results as all previous times it was computed. Deterministic models are sometimes compared to stochastic models where the same boundary conditions, initial conditions, and parameters are represented with probabilistic distributions. While not the same as stochastic models, the Uncertainty Analysis allows the calculation of probabilistic results using a Monte Carlo simulation of the deterministic models.

Most of the mathematical models included in the program use constant parameter values, that is, they are assumed to be time stationary. During long periods of time it is possible for parameters describing a watershed to change as the result of human or other processes at work in the watershed. Some model parameter can be varied in an annual pattern or adjustments made at specific points in time. There is a limited capability to break a long simulation into smaller segments and manually change parameters between segments. Plans are underway to develop a variable parameter capability, through an as yet undetermined means.

Key mathematical models representing the land surface are quasi-coupled. In the physical world, both the amount of evapotranspiration and the amount of infiltration each depend on the amount of soil water. However, evapotranspiration removes water from the soil at the same time infiltration adds water to the soil. Ideally all of the equations for the relevant processes would be solved simultaneously. Quasi-coupling is used in order to maintain flexibility in independently choosing the representation of each process. First, potential evapotranspiration is computed based on atmospheric conditions. Second, precipitation is added to the soil water. Third, actual evapotranspiration is computed based on the soil water content and then removed. Finally, a mass balance is performed and soil water exceeding capacity is returned to the land surface to become overland runoff. Any remaining errors due to the quasi-coupling scheme can be minimized by using a suitably short simulation time interval.

## Flow Representation

The design of the Basin Model only allows for dendritic stream networks. The best way to visualize a dendritic network is to imagine a tree. The main tree trunk, branches, and twigs correspond to the main river, tributaries, and headwater streams in a watershed. The key idea is that a stream does not separate into two streams. The basin model allows each hydrologic element to have only one downstream connection so it is not possible to split the outflow from an element into two different downstream elements. The 2D transform option does allow the modeler to simulate backwater, branching, and looping flow within a subbasin element's 2D grid. The diversion element provides a limited capability to remove some of the flow from a stream and divert it to a different location downstream in the network. Likewise, a reservoir element may have an auxiliary outlet. However, in general, branching or looping stream networks cannot be simulated with the program and will require a separate hydraulic model which can represent such networks.

The design of the process for computing a simulation does not allow for backwater in the stream network. For each time interval, the compute process begins at headwater subbasins and proceeds down through the network. The process starts with each subbasin computing outflow for the time interval. These outflow values are passed to downstream elements where they become inflow to that element. Continuing downstream, each element receives inflow from upstream elements for the time interval, and then computes its own outflow. Because outflow is only passed from upstream to downstream elements, it is not possible for an upstream element to have knowledge of downstream flow conditions, which is the essence of backwater effects. There is a limited capability to represent backwater if it is fully contained within a reach element. However, in general, the presence of backwater within the stream network will require a separate hydraulic model.