Dr. George Ice, Principal Scientist

Dr. Ray Whittemore, Principal Research Engineer

National Council of the Paper Industry for

Air and Stream Improvement

Water resource agencies have identified the need to develop watershed-scale assessments to evaluate progress in meeting the goals of the Clean Water Act. EPA’s Watershed Initiative exemplifies this new focus. The forestry community has been a leader in developing these watershed evaluation techniques. Three distinct approaches are discussed and examples are provided. These include large watershed-scale monitoring, watershed-scale adaptive management assessments, and watershed modeling/monitoring combinations.

Large watershed monitoring differs from site-specific or small watershed monitoring in the critical treatment of transport and fate monitoring. Often, these studies involve measurement of tributaries and multiple reach response.

Adaptive management approaches are designed to learn from ongoing management. One well-accepted adaptive management approach, Watershed Analysis, is a structured procedure for examining watershed conditions, landscape and management hazards, and aquatic resources at risk. Another adaptive management approach for watersheds is the Source Search Method. This can involve a synoptic survey to identify "hot spots" associated with specific management and site condition combinations.

One of the most appealing approaches is the development of realistic models. Models can be used to test different alternatives and are not confounded by the weather or watershed variability associated with even well-paired adjacent basins. The development of calibration and validation data sets is critical to making models effective. Some examples of watershed-scale models used in assessing water quality response include DHSVM, BOISED, and BASINS2.

These examples demonstrate that modeling and monitoring should be coordinated to efficiently assess BMPs at the watershed scale.

The Federal Clean Water Act and development of state nonpoint source control programs for silviculture necessitated evaluation of control practices. While early assessments focused on site-specific impacts and local response, more comprehensive assessments are now being required to assess the effectiveness of controls at a watershed scale and cumulatively with other activities. This paper provides a guide to approaches useful for evaluating the effectiveness of forest practices at the watershed scale.

Protection of watershed and water quality on forest lands predates passage of the Federal Water Pollution Control Act Amendments of 1972 (PL 92-500), commonly referred to as the Clean Water Act. Observations concerning impacts to streams from forest management activities go back to the birth of forestry in the United States (Schenck 1955). The origins of the National Forest system are tied directly to watershed concerns. Steen (1991), writing about the early legislation that established the National Forests, concluded that: "the primary driving force behind forest reserve legislation at that early time was the protection and enhancement of water supplies, including flood protection."

Forest watershed studies began early in the century (Bates and Henry 1928) and have continued and intensified at a number of well-known sites in this country (e.g., Coweeta Hydrologic Laboratory, Hubbard Brook Experimental Forest, H.J. Andrews Experimental Forest, and Caspar Creek Watershed). However, passage of the 1972 Act greatly accelerated the development of state water quality protection programs for forestry and the need to assess the effectiveness of control practices. First, the Act recognized the need for area-wide protection programs. Under Section 208, states were required to develop area-wide (watershed or regional) water quality management plans. Second, the Act recognized two classes of pollutants: point sources, which are "end-of-pipe" sources usually associated with industrial facilities or municipal waste treatment plants; and nonpoint sources, which are diffuse sources, often generated by hydrologic events, that are difficult to distinguish from background loads.

One step in developing area-wide water quality protection programs was to "…identify, if appropriate, agricultural and silvicultural nonpoint sources of pollution…" and to develop "…procedures and methods (including land-use requirements) to control, to the extent feasible, such sources" (Senate Committee on Environmental and Public Works 1978).

Originally, this area-wide planning focused on locations identified as having water quality problems. This approach was successfully challenged in court by the Natural Resource Defense Council (NRDC vs. Train). As a result, all state lands (not just designated areas) were subject to planning under Section 208.

In 1975, as required by NRDC vs. Train, EPA developed revised regulations to implement Section 208. Rey (1980) reported that the new regulations established the concept of Best Management Practices (BMPs) as the appropriate tool for nonpoint source control. EPA defined a BMP as:

The concept of "better land-management practices" for forest watersheds goes back at least 50 years (Craddock and Hursh 1949). However, the EPA regulations for the 208 programs solidified BMPs as the nonpoint source control tool of choice. O’Laughlin (1996) stated that:

Operationally, BMPs are useful for forest nonpoint sources because they are often designed to prevent adverse impacts before they occur, and they provide a measure of certainty about the operator’s responsibility to protect water quality. Also, it is generally easier to assess BMP implementation than it is to measure water quality standard compliance in dynamic forest stream systems. However, this means that state and federal agencies must connect BMPs with the goals of "…reducing the amount of pollution generated by nonpoint sources to a level compatible with water quality goals" (Rey 1980).

Initial assessments of BMP effectiveness involved research plots, field evaluations, small paired watershed studies, and application of agricultural models like the USLE to forest conditions. These all provide valuable information, but they are also limited. Most were designed to provide only a local assessment of impacts, often to an individual operation. The assumption was that if impacts could be minimized at the site, they would be diluted downstream. However, concerns about cumulative effects caused forest managers to explore alternative assessment approaches.

Cumulative watershed effects (CWE) for forestry are defined as "…changes to the environment caused by the interaction of natural ecosystem processes with the effects of two or more forest practices." Early assumptions about downstream dilution, transport of impacts, and stream response were challenged. Channel classification became an important factor in interpreting the potential for response to upstream management impacts. Federal agencies began analyzing for cumulative watershed effects from forestry under the National Environmental Policy Act of 1969 (NEPA) and requirements of the Clean Water Act (Coats and Miller 1981). Most of the early assessment methods used additive models that accumulated material loads to streams or indirect indexes of cumulative effects such as the equivalent roaded area method. In some National Forests, management limits were set, based on the percent of watershed harvested or roaded, to avoid cumulative effects. Often these limits were imposed without regard to site conditions, practices, or BMPs applied to control impacts. These types of limits continue today in proposals such as the Columbia Basin Intermountain Plan. With maturation of CWE assessments, watershed specialists began to recognize both the dynamic nature of watersheds and streams, and the potential for and need to address operational fall-down of BMPs (Callaham and DeVries 1987). This led to development of BMPs such as "diversion proof road designs" (Hagans and Weaver 1987) and "debris torrent-resistant road crossings" designed to minimize impacts to watersheds during extreme events.

Coincident with the rise in interest in CWE assessments, watershed management received increasing attention by the early 1990s. This attention was highlighted by the overwhelming interest in the Watershed ¢ 93 Conference and the growth of watershed-related organizations like the Watershed Management Council and the American Institute of Hydrology. In recent years, watershed-scale assessments were further stimulated by legal requirements to develop total maximum daily load limits (TMDLs) and the growth of GIS technology and landscape ecology methods capable of addressing spatially complex watershed problems.

Forest nonpoint source control programs are increasingly being asked to assess the effectiveness of BMPs at a watershed scale for multiple activities over long time periods to demonstrate that watershed, water quality, and landscape ecology goals can be met.

The Washington State Timber/Fish/Wildlife (TFW) Water Quality Monitoring Steering Committee report (1997) identifies long-term watershed trends as a key component to assessing BMP effectiveness. Basin-wide cumulative response to BMPs remains one of several "potential" issues for the forestry community in its efforts to validate the effectiveness of BMPs. But larger scales for both spatial and temporal assessments create problems for those attempting to validate BMPs. While this is a difficult task, there are three general approaches: adaptive management approaches, large watershed monitoring, and models.

Small watershed studies, upstream/downstream monitoring of management activities, and direct on-slope measures of BMP effectiveness provide for much more control of conditions and opportunities for replication of study sites. So why is it desirable to have a large watershed-scale component for monitoring effectiveness? There are at least five reasons. A watershed-scale assessment allows for an integrated or cumulative measure of BMP and program effectiveness. It allows BMP effectiveness to be placed in the context of realistic variations in water quality throughout a watershed (see "source search" discussion in Adaptive Management Approaches section) and over time; it allows for assessment of the conservative/non-conservative nature of water quality parameters; it connects upslope hazards with downslope aquatic resource risks; and it allows for assessment of unanticipated consequences that might not be identified at the site scale. We can show that overall water quality has improved for the nation since the 1972 Federal Water Pollution Control Act Amendments were enacted. We can document that silvicultural BMPs have the capability to protect water quality. But can we document that overall water quality is benefiting from forest NPS control programs?

One other important reason for looking at effectiveness on a watershed scale is new information about what scales are most sensitive to management. While the smallest headwater streams can have the greatest changes because impacts are not diluted and load-to-flow ratios are the greatest, these systems are also the most variable naturally. Often, headwater systems are exposed to dramatic changes in flow and water quality even without management impacts and organisms are adapted to these fluctuations. In large streams, the non-conservative nature of pollutants causes upstream impacts to be muted as the system moves toward equilibrium with its environment. Sensitive reaches, such as unconfined channels or deposition zones, may occur at key locations in the watershed, and risks to these sites need to be linked with upslope hazards.

Adaptive management approaches are designed to utilize ongoing management as a test from which to learn. Probably the two most useful examples of adaptive management monitoring at the watershed scale are source searches and watershed analysis.

Source search or synoptic survey assessments involve a snapshot of water quality throughout a basin at one time. While there are potential problems associated with this type of infrequent monitoring view (i.e., potential differences in hydrologic conditions in the watershed or unrepresentative conditions for estimating pollutant loads), this can be a powerful tool. An example is the Mokelumne River Watershed in the Sierra Nevada Mountains of California, where accelerated forest management activities in the watershed were blamed for eutrophication of downstream reservoirs. A synoptic survey found that 5% of the watershed (below the portion of the watershed experiencing forest management) was contributing 60% of the nitrate load (Dahlgren 1996). This was found to be a result of geologic sources. Annual variations in eutrophication were found to correlate better to basin water yield and lake residence time (well-known factors affecting eutrophication).

Watershed Analysis (WA) is "…a structured approach to develop a forest practice plan for a [Watershed Administrative Unit] based on a biological and physical inventory" (Washington Forest Practices Board 1993). Information about watershed hazards from forest management and public resources at risk are used to develop watershed-specific prescriptions (Figure 1). For example, there may be evidence of past erosion in the watershed associated with roads constructed on erosive soils near streams without proper control of runoff. Sediment transported downstream from these sediment sources may have resulted in the filling of pools and degradation of habitat quality for fish. These types of observations result in "conditioning" of BMPs to prohibit or place management limits on roads constructed in the landtypes and conditions found to cause problems. WA has some very structured elements, such as:

• minimum education and training requirements for WA team members
• standard assessment methodologies
• module synthesis by the expert teams
• management and public involvement
• rewards for participation
• situational syntax and hypothesis development
• monitoring of performance and use of triggers for reassessment

Watershed Analysis provides useful stand-alone watershed scale information by developing hypotheses that connect upslope hazards with downstream risks through key watershed processes. It has stimulated development of basin-scale assessment tools. It has also allowed sharing of experiences between watershed assessments.

How does a large watershed monitoring study differ from a small basin study? Obviously the scale of the watershed differs, but the questions to be addressed and the monitoring design also differ to some extent. Small basin studies usually measure the integrated response to management and may provide some on-slope or upstream monitoring to support the assessments about impacts. Large watershed studies are usually designed to measure tributary and multiple reach response to assess the conservative/non-conservative nature of pollutant/flow transport. This can allow an interpretation of the relative hazards in the watershed (erosive soils), risks (i.e., deposition zones), and connections. One of the first examples of this type of large watershed study was Caspar Creek in California, where multiple subbasins and mid-reach sections were monitored (Cafferata 1984). Another example is the Mica Creek watershed study in Idaho (McGreer et al. 1995). For Mica Creek, watersheds are nested to allow assessment of impacts as they are routed downstream. Watershed 3 (613 acres) in the headwaters of the West Fork of Mica Creek serves as a control to adjacent Watersheds 1 (354 acre) and 2 (440 acres). Watershed 5 on Mica Creek (1,655 acres) serves as a control for the watershed formed below 1, 2, and 3 at Station 4 (1,457). Station 6 downstream on Mica Creek (3,616 acres) is a control for Station 7 (3,031) on the West Fork of Mica Creek (Figure 2).

The Alabama Demonstration Watershed Project provides another example of a large basin monitoring effort. This effort involves a screening of 2,500 to 12,000 acre (4 to 18 mile²) subbasins within the Sepulga River Basin of southern Alabama (Lockaby et al. 1996). The total Perdido-Escambia River Basin, including the Sepulga Basin, covers over 5,000 square miles. Comparisons are being made between land-use activities, BMP implementation, and water quality for the subbasins. More detailed monitoring of subbasins, including source search for pollution-generating sites, will occur as contrasting land-management history/water quality subbasins are identified.

One other kind of large watershed scale monitoring approach takes advantage of remote sensing opportunities to collect information where site-specific information is not useful or where technology provides advantages for basin-wide assessments. Aerial reconnaissance methods are used to assess forest management impacts that are often widely distributed but observable from the air. For example, landslides are traditionally measured at the watershed or even regional level using aerial photos or aerial reconnaissance. Recent work by the Oregon Department of Forestry is providing some information about the detection bias associated with aerial photo and aerial reconnaissance methods compared with on-the-ground measurements (Robison 1996). NCASI is working with researchers at Oregon State University to develop a low cost, rapid response video method for recording landslides (Rosenfeld et al. 1996). Thermal remote sensing has also been used to track stream temperatures quickly throughout a basin (Torgersen 1996).

NASA has recently released a Research Announcement to "...solicit proposals for the establishment of Regional Earth Science Application Centers (RESACs) designed to apply remote sensing and attending technologies to well-defined problems and issues of regional significance." Real-time use of remote sensing information by water resource managers is one of the sector-specific problems identified. The application of large-scale remote sensing information to water quantity and quality assessments could present a major advance in watershed assessment.

One of the major problems with developing large paired basin comparisons is that even small adjacent watersheds can experience dramatically different weather patterns. One of the approaches that is most appealing to foresters is the development of realistic models that can simulate response to forest management practices at both the site and watershed scale. This approach allows forester managers to predict the impact of their actions on water quality and develop management solutions, rather than just documenting impacts after the fact. Unfortunately, much work has gone into developing various watershed models, while little work has gone into validating these models, calibrating them to different watershed conditions, and making them user-friendly. One example is the effort to apply the USLE to forest watershed assessments (James and Hewlett 1992). Conversion of GIS data was found to create major modeling errors and overestimates of erosion and sedimentation.

NCASI has been working with the University of Washington and Battelle Northwest Laboratories to validate the Distributed Hydrologic-Soil-Vegetation Model (DHSVM) against watershed data collected by companies for different conditions. "DHSVM accounts explicitly for the effect of topography and the spatial distribution of land surface processes at the scale of currently available Digital Elevation Models (30-90 m)" (Wigmosta 1996). Features include spatially distributed, digital elevation model grid-based approach; automated model setup using the GIS ARCINFO; explicit, spatially-distributed representation of road networks; spatially distributed vegetation and soils properties; topographic control on absorbed short-wave radiation, precipitation, and downslope water movement; a two-layer soil rooting zone model; a spatially distributed two-canopy evapotranspiration model; simplified topographically-driven surface and subsurface flow routing; GIS post-processing of model outputs; and channel flow routing (Figure 3). Testing of DHSVM is occurring on gaged industry watersheds, including the Little Naches in central Washington, the Deschutes in western Washington, Mica Creek in northern Idaho, and Carnation Creek on Vancouver Island (Figure 4). This model has been converted from requiring a mainframe computer to a PC-based Windows NT environment. A users manual is planned for development in 1998.

A very useful application of a well-calibrated empirical model is provide by Megahan et al. (1992). Megahan et al. used BOISED, a version of the Forest Service R1-R4 sediment yield prediction model (WATSED), to provide a retrospective assessment of sediment yield for a tributary of the South Fork of the Salmon River. WATSED is based on locally derived empirical streamflow and sediment yield data, and uses stand properties and landscape units defined in terms of landform, lithology, and soil characteristics. Onsite surface and mass erosion estimates are adjusted for slope delivery based on topographic conditions, and downstream sediment delivery is adjusted on the basis of a watershed sediment delivery ratio. The model is sensitive to alternative forest cutting and soil disturbance activities, including silvicultural practices, alternative road construction practices, and wildfire. Megahan et al. estimated that abusive jammer logging and associated road construction boosted sediment yields from about 450 Mg of sediment to over 1300 Mg. With present day BMPs, the authors estimate sediment increases could have been reduced by 45 to 95% (Figure 5).

An emerging tool for modeling is BASINS2 (Whittemore and Ice 1998). BASINS2 is a comprehensive EPA software package recently released by the Office of Water. It is designed to enable water quality analysts and watershed managers to perform studies using geographic information system (ArcView), watershed landuse and water quality monitoring data, and state-of-the-art environmental assessment tools. BASINS2 provides information for any of the 2,150 watersheds in the conterminous United States. It incorporates models such as the Nonpoint Source Model (HSPF version 11), TOXIROUTE, and QUAL2E. BASINS2 has much promise, but its coarse spatial scale and treatment of land-use activities do not support assessments of alternative forest management activities at this time (Figure 6).

Clearly, modeling and monitoring are inseparable (EPA 1997). Even detailed process models require calibration and validation studies to ensure that they are performing acceptably. In the future, we may begin to develop hybrid monitoring and modeling systems which incorporate real-time remote sensing data to parameterize models and provide for rapid calibration to existing conditions. These models can then be used to test alternative management scenarios. In the interim we must continue to use the results of past experience along with careful translations of those experiences to the site- and watershed-specific conditions encountered.

Figure 3. DHSVM model utilizes GIS data to simulate spatially explicit hydrologic response and route discharge through road and stream reaches.

Figure 4. A comparison between observed discharge for The Little Naches River in central Washington and simulated discharge using DHSVM (from Wetherbee and Lettenmaier in press).

Figure 5. Megahan et al. (1992) used WATSED to demonstrate how BMPs have reduced sediment
yield from historic loads in Dollar Creek, ID.

Figure 6. BASINS2 combines easy access to data, integrated watershed models,
and a PC-compatible Windows environment.