Most large nonlinear optimization models are composed of "pieces" -subsets
of decision variables-and constraints whose union is the entire model. Each
piece represents an additional aspect of the situation being modeled. This
opens the possibility, of solving the simplest piece first, adding the con
straints and variables of another piece, and solving this submodel from a s
tarting point provided by the first solution. This process is repeated unti
l the original model is solved. This "piece-by-piece" approach provides eac
h submodel with a good starting point, which greatly increases the probabil
ity that a good nonlinear solver will find an optimal solution. We apply it
to a large multiperiod nonlinear programming (NLP) model with 13,700 varia
bles, 10,000 equations, and a high degree of nonlinearity (54.3% of the non
zero Jacobian elements are nonconstant), arising from water resources plann
ing and operation in. a river basin. Using the GAMS modeling language and t
he CONOPT2 NLP solver, the piece-by-piece method is able to solve this mode
l, while all attempts to solve the complete model from various starting poi
nts fail to find a feasible solution.