species occur; large water bodies such as the Chesapeake Bay; and many, many others.

Forest hydrology is adopting a landscape perspective to examine spatial patterns of forests and associated hydrologic processes and to link principles from plot- and small watershed scales (up to several square kilometers, see Chapter 3) to predictions at larger spatial scales (hundreds of square kilometers). Within a watershed, forests can be located in the headwaters, along riparian corridors, in woodlots in agriculture lands, and in urban or suburban areas (Figure 4-1). Based on its intra-basin position, a forest fulfills various water-related functions with respect to water quantity and quality. For example, forests in headwaters influence water yield and the quality of water delivered to downstream areas. Riparian forests located along streams throughout a watershed provide key functions for protecting streams from inputs of sediment (Naiman and Decamps, 1997), nutrients, and herbicides (Peterjohn and Correll, 1984; Lowrance et al., 1997); provide wildlife habitat for terrestrial and aquatic organisms (Barton et al., 1985; Darveau et al., 1995); and support a diversity of other functions (Risser, 1995). Riparian forests have been greatly altered by economic development, and they are the focus of many forest management guidelines designed to preserve water quantity and water quality.

Research Need: Process studies are needed to determine how forests, particularly riparian forests, affect water quantity and quality according to their position within a watershed.

Hydrologists use models to predict water quantity and quality in watersheds where there are no measured records. Since 2004, a working group for Prediction in Ungauged Basins (PUB) of the International Association of Hydrological Sciences (http://www.hydrologic science.org/pub/about.html) has developed methodologies for assessing uncertainty in hydrologic predictions arising from uncertainties in landscape properties and climate inputs, choice of model structure, and methods of information transfer from gauged to ungauged watersheds (Sivapalan, 2003). Most hydrological models are developed and tested for gauged basins and subsequently are validated and applied to ungauged areas. However, models that have been fitted to data in small, gauged watersheds often provide inaccurate or imprecise predictions when they are (1) extrapolated to other small forested headwater basins, (2) extrapolated to future time periods, or (3) applied to large watersheds. This problem of prediction in ungauged basins has preoccupied hydrology researchers for several decades, and is compounded by a lack of information about how direct hydrologic effects interact under the multiple sets of specific conditions that occur in changing forest landscapes. By examining forest hydrologic processes under a wide range of conditions, landscape-scale studies could provide data and understanding to help extend basic forest hydrology principles to make predictions needed by water managers.