ABSTRACT: Spatially-distributed depth and velocity predictions are required for habitat based instream flow studies. The purpose of this thesis is to estimate uncertainty of two-dimensional (2D) depth-averaged hydraulic models when applied with close spacing of computational nodes. Motivation for close node spacing is discussed from the ecological, aquatic habitat perspective. Model-generated maps of predicted depth and velocity require sufficient resolution to capture spatial variations relevant to aquatic habitat; however, bathymetric variations at that resolution are more complex than strictly applicable for the depth-averaged hydrostatic model equations. Hydraulic model assumptions are discussed and the geometry of a typical model is analyzed to identify areas that do not conform to assumptions. Model input data, including bathymetry, water surface elevation, flow rate, depth and velocity measurements, have accuracy within 5% of actual values. Accuracy of depth measurements conducted with a boat-mounted echosounder approach 15 centimeters and are the greatest source of uncertainty for depth error in model predictions. For model test scenarios using the RMA2 2D depth-averaged finite element code, geometries exhibiting slopes greater than 0.10 (ratio of rise to run) or exhibiting abrupt lateral changes in width are shown to cause changes in continuity (velocity conservation) of greater than 2.5%. For a calibrated model of the Brazos River, Texas, 95% of the model area exhibited low uncertainty with continuity deviations less than 2.5%; remaining areas exhibited higher uncertainty resulting from steep slopes or high Froude numbers.