The leakage characteristics of channels that partially penetrate the Biscayne aquifer and have reduced bed permeability were studied. Leakage characteristics were described in terms of a reach transmissivity-defined as the volume flow rate out of the channel per unit length of the channel per unit drawdown, where drawdown is defined as the difference in altitude between the water surface in the canal and the water table in the adjacent aquifer. A theoretical expression was developed to relate the reach transmissivity to the transmissivity of the formation, mean channel width, distance of drawdown measurement from the channel centerline, ratio of drawdowns on both sides of the channel, and local reach transmissivity associated with reduced bed permeability. This theoretical expression was verified using a fine-scale numerical model, which gave accurate results when drawdowns were measured beyond 10 aquifer depths from the side of the channel. Using the theoretical formulation, it is shown that the reach transmissivity employed in regional ground-water models, which are based on average drawdowns within a cell, depends on the size of the cell as well as the transmissivity of the formation, channel width, and local reach transmissivity due to reduced bed permeability. The theoretical reach transmissivity function was compared with field measurements at L-31N Canal and Snapper Creek Extension Canal in Dade County, Florida. Analyses of the data for both canals showed good agreement between the estimated and measured reach transmissivities. At L- 31N Canal, field measurements indicated that the local reach transmissivity was relatively uniform over a 2-mile reach of the channel (averaging 630 cubic feet per second per mile per foot), and the formation transmissivity was 1.8 x106 feet squared per day. At Snapper Creek Extension Canal, an approximate analysis was necessary due to the inability of the acoustic velocity meter to measure very low water velocities in the channel. Assuming an aquifer transmissivity of 1 x 106 feet squared per day, drawdown measurements indicated that the local reach transmissivity was about 400 cubic feet per second per mile per foot. The theoretical relation, combined with the local reach transmissivity and formation transmissivity, was sufficient to predict the leakage out of L-31N Canal and Snapper Creek Extension Canal for any drawdown scenario.