We use a nonlinear least squares inversion to derive predicted seafloor topography (hereinafter referred to as RW99) for the New Zealand region (146 ¿E--165 ¿W, 60 ¿S--25 ¿S), combining altimetry data from ERS-1 and Geosat Geodetic Missions, as well as available shipborne gravity and echo sounding data. Currently, the lithospheric component of the model is principally applicable to thinly sedimented oceanic basins; however, we have attempted, though with only partial success, to compensate for regional crustal variations. The upper part of the oceanic lithosphere has an elastic behavior related to a half-space cooling model and flexing under seafloor relief load. Using least squares theory, the topographic solution is derived as a linear combination of altimetry and in situ measurements with adjusted coefficients. These coefficients are iteratively fitted using nonlinear operators between bathymetry and altimetry-derived gravity anomalies assuming their error distributions are Gaussian. The theory enables sparse in situ data to be included in the inversion, such as depth soundings and marine gravity profiles. In comparison with the global model of Smith and Sandwell <1997> (hereinafter referred to as SS97), the RW99 predicted topography is constrained by over threefold more shipborne soundings data and the inclusion of shipborne gravity data. Three strategies are used to validate the RW99 model. Compared to the root-mean-square (rms) error of 310 m of the SS97 model, final residual differences for the RW99 model are within the range of 104--250 m. These rms errors are the result of uncertainties of model parameters, especially the elastic thickness and the relief density, but also the complexity of seafloor topography. In addition, the model inversion does not presently consider gravitational contributions of marine sediments of variable thickness.