A deep percolation model, which operates on a daily basis, has been developed to estimate long-term average groundwater recharge from precipitation. It has been designed primarily to simulate recharge in large areas with variable weather, soils, and land uses, but it can also be used at any scale. This report documents the physical and mathematical concepts of the deep percolation model, describes its subroutines and data requirements, and describes the input data sequence and formats. The physical processes simulated are soil-moisture accumulation, evaporation from bare soil, plant transpiration, surface-water runoff, snow accumulation and melt, and accumulation and evaporation of intercepted precipitation. The minimum data sets for the operation of the model are daily values of precipitation and maximum and minimum air temperature, soil thickness and available water capacity, soil texture, and land-use. Long-term average annual precipitation, actual daily stream-discharge, monthly estimates of base flow, Soil Conservation Service surface-runoff curve numbers, land-surface altitude-slope-aspect, and temperature lapse rates are optional. The program is written in the FORTRAN 77 language with no enhancements and should run on most computer systems without modifications. Documentation has been prepared so that program modifications may be made for inclusions of additional physical processes or deletion of ones not considered important.