How are the NARR topography and land/water masks created?
The NARR topography and land/water masks are created in the same way as those of the then operational Eta model, following the procedure as follows. Each model grid box is split into 2 x 2 subboxes, and terrain elevation read off terrain data (USGS 30 s data where available), collecting values and percentages of water that happen to fall within each of the subboxes. Using these, mean subbox elevation and percentage of water is obtained. Subsequently, for each grid box, mean and silhouette elevations are calculated. Silhouette elevation is calculated by looking at each pair of subboxes as seen from two directions perpendicular to sides of the box, amounting to four pairs total, each time taking the higher value, and subsequently averaging these four higher values. Following this, nine-point Laplacian of the mean orography is calculated for every box. Where the Laplacian is positive, mean is used for the box elevation. Where it is negative, silhouette is used (Mesinger, Bull. Amer. Meteo
