2008 Progress Report: Quantifying Stream Ecosystem Resilence To Identify Thresholds For Salmon Recovery

EPA Grant Number: R832439
Title: Quantifying Stream Ecosystem Resilence To Identify Thresholds For Salmon Recovery
Investigators: Merenlender, Adina , Kondolf, Matt , Moyle, Peter , Resh, Vincent
Institution: University of California - Berkeley
EPA Project Officer: Packard, Benjamin H
Project Period: July 1, 2006 through June 30, 2007 (Extended to August 31, 2008)
Project Period Covered by this Report: July 1, 2005 through August 30,2008
Project Amount: $299,922
RFA: Exploratory Research: Understanding Ecological Thresholds In Aquatic Systems Through Retrospective Analysis (2004) RFA Text |  Recipients Lists
Research Category: Aquatic Ecosystems , Ecosystems , Water


The distinct wet and dry seasons found in mediterranean-climate regions such as northern coastal California results in stream flow receding to approach or reach intermittency through spring and summer. The dry season can put stress on macroinvertebrates and native anadromous salmonids that use freshwater streams as habitat for their juvenile life cycle stages. Intensive agriculture in these areas requires water during the warm growing season. The cumulative effects of removing water for agricultural use during the dry season can result in stream levels dropping below critical ecological thresholds, thus increasing the probability of salmonid and macroinvertebrate mortality. To quantify the human impacts on the natural hydrologic regime and identify resilience thresholds we proposed the following methods: (I) creating models of natural flow regime on a daily-scale by using historical stream flow and rainfall data; (II) creating spatially explicit estimates of water demand in each stream in the study region over time; (III) identifying reaches where stream flow may be adversely affected by human water extraction and evaluating trade-offs among water management options; and (IV) exploring the potential impacts that spring and summer stream flow dynamics have had on stream biota, evaluated using long-term fish and macroinvertebrate monitoring data. The data available for quantifying the variables described above are not perfect, but sufficient data do exist for estimating when human demand may lower resilience thresholds, leading to alternative states of biotic interactions in small coastal California streams.

We tried to identify streamflow thresholds that, when crossed, lead to changes in macroinvertebrate community structure and salmonid persistence/survivorship. Aquatic communities in our study region must withstand the drought conditions that occur each summer (wherein flows typically approach or reach intermittence) to persist in these environs; but particularly harsh conditions driven by prolonged drought or water removal for agricultural and other uses may cause re-structuring of ecosystems to an alternative regime. These alternate regimes may manifest themselves as a different macroinvertebrate community than would typically be expected, or decreases in anadromous salmonids in a particular reach or drainage.

Progress Summary:

During the past two years of this research, we have made progress on all four tasks listed above for several upstream tributaries in North Coast California.  We are studying stream ecosystems in tributaries to the Russian River and Putah Creek in northern coastal California.  These streams provide habitat for a complex and diverse macroinvertebrate community throughout the drainage network; and anadromous salmonids use the Russian River tributaries for spawning in winter and juvenile rearing in summer.  A persistent low-flow period from late spring until early fall results from the absence of precipitation between May and October; rainfall is highly variable across space and between years.  Streams also represent an intersection of human-ecosystem interactions: wine grape growers in the region use these streams as a source for water needs during the growing season, which may amplify or accelerate the summer drought conditions. Therefore, we provide modeled stream flow and calculate agricultural water demand throughout the study watersheds in the Russian River; and use this information to evaluate the cumulative impact of small distributed reservoirs.

I. Modeling upland stream flow

Two types of flow modeling have been employed to address the primary research objectives of this study.  The first simply uses streamflow data at centrally located USGS gauge stations that had greater than 10 years of records, scaled these streamflow data by watershed area and mean annual precipitation to create a daily streamflow value for each point (every 10m) in the drainage network.  This proved to provide better estimates of low flow dynamics than the methods described in our original proposal which involved using a statistical model of stream flow as a function of rainfall for the day, as well as rainfall over previous antecedent periods using a multiple regression analysis (see 2nd annual report).  In that case, the USGS measured flow in Maacama Creek and Santa Rosa Creek (tributaries to the Russian River in eastern Sonoma County, CA) during the 1960s and 1970s, and we used precipitation records at nearby weather stations to derive regression models to predict mean daily flow as a function of rainfall over various antecedent periods.  The strongest statistical relationships were derived when time was separated into spring (March through June) and summer (July through September).  For both streams, we derived multiple regressions from historical stream flow and precipitation data where spring flow is a function of the day’s precipitation, precipitation over the previous two days, precipitation 3 to 7 days before, precipitation 8 to 30 days before, and precipitaiton 31 to 90 days before; and where summer flow is a function of the last 3 days’ precipitation, precipitation 4 to 30 days before, precipitation 31 to 90 days before, and precipitation 90 to 210 days before.  Because flows were recorded over 21 years in Maacama Creek (from 1961 to 1981), a sufficient number of years’ flows were available to create the model and to test it as well. Two-thirds of the years when flows were measured in Maacama Creek were randomly selected to create the spring and summer models; the other third were used to test the models.  Simple flow-scaling techniques are also employed to estimate flow at sites along the same stream. Data indicate that the models created a reliable estimate of flows in winter and spring but worked less well for the dry summer period (Figure 1).  This can be observed by the flatter flow curve predicted over the summer period that is missing the empirically observed fluctuations in low flow. 

II. Water demand estimates

A vineyard map based orthorectified aerial photos from 1993, 2000, 2002, 2004 and 2005, as well as oblique aerial photos from 2006 (59,000 acres in Sonoma County; for more details see research update “Mapping shows continued vineyard expansion in premium wine-growing areas” California Agriculture January-March 2008 and Merenlender 2000) was used to estimate the agriculture water need by multiplying each acre by 2/3 acre foot (this estimate does not include additional water needed in areas where overhead sprinkling is required for frost protection, thus making ours a conservative estimate of the water needs during the spring). Added to this need is an additional 0.226 acre-feet per rural residential unit to account for outdoor water use by the average home. Rural residential units were mapped based on the County parcel and assessor’s data that includes units per parcel.  The estimated water need was then summarized for each individual land parcel.

Reservoirs were also digitally mapped from aerial photographs and the surface area for each reservoir was used to estimate total volume based on a empirical statistical relationship between a sample of recorded volumes and surface area (N=100) from the SWRCB.  The estimated volume of winter water storage in existing reservoirs on a given parcel was subtracted from the estimated water need per parcel described above.  We then created centroids of each parcel that contained the attributes of total water need per parcel not met by winter water storage.  Total water need for each land parcel was then accumulated using a flow accumulation model (ESRI 2006) to determine the cumulative need through the entire drainage network.  Figure 2 shows an example of these estimates mapped cumulatively downstream for two adjacent watersheds within the Russian River Basin.   

III. Evaluating trade-offs between water management options and restoration planning

Surface water diversions may have the most substantial impacts on aquatic biota during the spring and summer because streamflow is naturally low: the limited water available is critical for maintaining suitable habitat conditions, yet streamflow at this time is most susceptible to impacts by diversions.  In many parts of the Russian River Basin, water rights records predict that demand for water during the spring and summer growing season exceeds supply, underscoring the imbalance between water need and supply during the growing season (Deitch et al., in press); yet normal-year discharge during the wet season exceeds annual water removal estimates by an order of magnitude.  In watersheds where water demand is high, the cumulative impacts of surface water diversions have the potential to accelerate drying over substantial stream reaches, reducing habitat availability for juvenile salmon and other aquatic species.  Secondary effects of stream drying, such as increased competition, higher water temperatures, and increased predation risk may also occur where flows are reduced (Kocker et al. 2008).

Since 1990, the State has hesitated to grant water rights in part because of uncertainties over whether additional appropriations will have cumulative effects on instream flows necessary to sustain salmonid migration, resulting in a backlog of requests for additional appropriative rights, many to increase the storage of winter runoff (SWRCB 1997; SWRCB 2007). Until recently, the basis for these decisions hinged on joint guidelines from California Fish and Game and National Oceanic and Atmospheric Administration (NOAA) Fisheries Service on maintaining winter flows sufficient for adult salmonid migration.  In December 2007, the State Water Resources Control Board (SWRCB) proposed regulations for surface water abstractions in northern coastal California related to aquatic ecosystem conservation, as part of its draft Policy for Maintaining Instream Flows in Northern California Streams (SWRCB 2007). 

Because the new policies for surface water appropriations may not allow grape growers to meet agricultural water needs, we expect that they will continue to turn to alternative means, including riparian water diversions and groundwater pumping in the growing season, neither of which are subject to the same standards as appropriations (Sax 2002), but which would likely have adverse consequences on stream biota because of their effects on naturally low spring and summer streamflow.  We theorize that it may be more useful to consider impacts of small water projects relative to local and cumulative impacts to discharge through the year, rather than to set a flow condition as the standard for diversion because it would preserve a particular flow condition.

To improve water management planning, ensure adequate flows for salmonid recovery, and avoid continued environmental and social crises over limited water supplies, we must integrate an improved understanding of ecological and human water needs at a fine spatial and temporal scale to accommodate the high levels of variability in water availability observed over space and time.  So we developed a spatially explicit model for agricultural and rural residential water need (parcel scale), daily stream flow throughout the watershed, proposed environmental stream flow requirements, and cumulative impact analysis of small reservoirs on streamflow.   Integrating this information at a fine scale across entire watersheds is essential to evaluate environmental and social tradeoffs associated with different water management schemes widely implemented across coastal California, where water for agriculture is not provided by large centrally controlled reservoirs. 

We created a model using our GIS to examine the cumulative impact of small surface reservoirs on stream flow through the year, as reservoirs fill from the onset of the rainy season in fall. Estimated reservoir volumes in the Sonoma County portion of the Russian River watershed (see above) were incorporated into our watershed model, and the upstream catchment area was calculated for each reservoir.  We modified the digital elevation model with inserted reservoirs so that water flowed from the upper watershed into the reservoirs until they filled and then out the lowest point of the reservoir into the downstream drainage network.  The start of the delineated network began at the reservoir outlets.  All segments of the stream network had the maximum flow accumulation value from upstream assigned and the hydrologic network was then exported from ArcGIS (ESRI 2006) as a vector line and point shapefile.  The data base files that were related to the shapefiles were then imported into a Microsoft Access database, where the database manipulated the stream network created by the GIS. The database was then used toaccumulate the estimated flow from the upstream drainage network across the grid cells within a watershed; the model assumes that reservoirs are empty at the onset of the water year, and that small dams block discharge from upstream until the reservoir fills  (i.e., when the cumulative discharge volume from the upstream watershed equals the volume of the reservoir), at which time the upstream drainage network is reconnected hydrologically with the rest of the watershed. 

We then used the flow accumulation model to determine the fraction of discharge accumulating from unimpeded parts of the watershed, and adjusted this fraction to reflect flow conditions as reservoirs fill through the winter.  In addition to showing local effects of reservoirs (i.e., immediately below the dam), the model is designed to illustrate the cumulative impacts of reservoirs on streamflow anywhere in the drainage network using the flow accumulation from the unimpaired drainage, considering the reservoirs upstream and the portion of watershed above each reservoir (Figure 3). 

Using normal-year flow data from a time of few dams and diversions, (thus representing unimpaired flow), the model indicates that early-season streamflow in some of the major tributaries to the Russian River may be reduced by as much as 50% and these impaired sites are predominately found in small watersheds.  Therefore, we expect that early-season rains may produce only a fraction of the streamflow that would be expected in the absence of small reservoirs.  However, the impact diminishes as the rainy season progresses because reservoirs fill over time: streamflow is reduced by less than 10% by the end of December for most reaches in normal rainfall years because many reservoirs have filled by this point.  Also, 90% of the highest impaired sites are in very small watersheds   Because the reservoirs common in the Russian River watershed are focused in headwater streams and the window for upstream bypass is larger lower in the watershed as compared to upper tributaries, these reservoirs are less likely to impact the ability for salmon to migrate through lower reaches to find suitable spawning tributaries. This modeling effort can help reveal where additional reservoirs for storing winter rainfall may be placed to minimize the impact to adult salmon passage and relieve the impacts of current management practices on spring and summer streamflow.

Applications & Next Steps

Despite the complexity and multiplicity of natural and anthropogenic stressors on river ecosystems in mediterranean climates, it is possible to reduce ecological impacts and improve our management of water resources to meet both human and ecosystems needs. The model we propose supports an integrated approach to water management by accounting for the spatial and temporal variability in water availability, human water needs and environmental flow requirements. In addition, the model allows for cumulative impacts analysis, which is often difficult to quantify but may be significant cause of ecosystem degradation in decentralized, water management systems. Furthermore, the approach can help to prioritize freshwater conservation efforts by evaluating the impacts and benefits of changes in water management practices on environmental flows. Finally, this approach is makes it possible to assess the consequences of alternative policy scenarios and supports integrated decision-making by institutions responsible for water and freshwater ecosystem management.

Our model is focused on management of surface flow in rivers and streams because in mediterranean regions they are a critical limiting resource for meeting both human water needs and sustaining ecological functions. However, in some locations groundwater is also important for meeting water needs and the extraction of groundwater has the potential to reduce surface flows and impact stream biota (Spina et al., 2006; Dewson et al., 2007). We expect future iterations of the model to incorporate surface-water groundwater interactions to improve our predictions of stream flow and water availability. We also plan to incorporate additional complexity in the model by considering other drivers of human water use practices, including water rights, land values, and local site topography.

Ultimately, the effectiveness of this model as a decision-support tool will be largely determined by institutional capacity to conduct analysis and develop management strategies at appropriate scales. This requires a formalized, integrated decision-making process and legitimate legal/political authority that are deficient both in mediterranean and non-mediterranean countries. Coordinated cross-agency efforts will be needed to conduct catchment-scale assessments and more importantly to implement resulting planning priorities. Moreover, land owner participation and support will be critical for the success of this coupled human-and-natural-systems approach to water management

IV. Crossing biological thresholds (macroinvertebrates and salmonids)

Study location

Our studies have focused on two particular areas in coastal California.  Macroinvertebrate and fish data were collected in Hunting and Knoxville Creeks in Napa and Lake Counties (I); and salmonid data were collected in 10 Russian River tributaries in Sonoma County (II).  See map of California with study sites (Figure 4) along with detailed study site maps (Figures 5 and 6).

  1. Napa and Lake County study

This study was a retrospective analysis of a long-term data set of invertebrates and fish sampled from 1984-2002 annually at four sites in Knoxville (site K1) and Hunting (sites H1, H2, H3) Creeks, in the University of California Donald and Sylvia McLaughlin Natural Reserve (Napa and Lake Co.), California, USA (Fig. 1).  These are small upland streams (watershed area: 2.1 to 29.3 km²) and summer flow is either reduced but perennial (H2 and H3), intermittent (disconnected pools; H1), or dry (K1).  We identified the duration and intensity of the drought period using data (1938-2004) from a nearby weather station (Angwin, California, USA).  Severe drought and wet years were identified as years within the 10th (< 550 mm) or 90th percentile (> 1500 mm), respectively.  Using these criteria, the five-year period from 1987-1991 and 1994 were identified as drought years (n = 6); 1987, 1990, and 1994 were classified as severe drought. 

We used repeated measures analysis of variance (ANOVA) to assess the effect of site, time, precipitation (wet, dry, and average years), and drought period (before, during, after) on invertebrate richness (number of taxa) and ln-transformed abundance.  Three models were examined: (1) Site*Year, (2) Site*DroughtPeriod + Year within DroughtPeriod, and (3) Site*Precipitation + Year(Precipitation).  To analyze changes in invertebrate composition, we performed three sets of multivariate analyses: (1) principal components analysis (PCA), (2) between-PCA, (3) multiple correspondence analysis (MCOA), and (4) repeated measures analysis of variance (ANOVA). 

Future Activities:

Over the period of record, we observed high variability in fish counts both temporally and spatially. Despite the high variability, the long term record make it possible to assess how the various flow, habitat and landuse variables explain the observed patterns in both initial counts and over-summer survival.

Median spring flows observed at the study reaches did not explain variation in juvenile salmonid recruitment, but was a significant predictor of over-summer survival. The regression model indicates that an increase of median spring flows by 10 percent is associated with a 26 percent increase in over-summer survival, averaged across all sites. The relationship between flows and fish survival did not appear to conform to a threshold pattern that was consistent across all sites. Rather, flow generally had a linear positive relationship to juvenile survival across the range of observed conditions. Our findings indicate that stream flow is critical to the survival of rearing juvenile salmonids during the dry season and suggest that changes in water management practices and restoration programs to increase flows could improve fish survival, even in streams that are highly impaired.

Despite the high variability in juvenile salmonid recruitment and survival between sites and years, the long-term record make it possible to assess quantify the effects of several relevant environmental variables operating at different temporal and spatial scales. Agricultural and land use development in proximity to the sampling reaches have negative effects on both recruitment and survival. The vineyard cover and road density variables used in the model likely served as proxies for other disturbances, such as sediment delivery and water diversions, but the findings nevertheless consistent with the well-documented adverse effects of land use on aquatic ecosystems. The regression models reveal (see Table 1)  that juvenile fish survival is highly sensitive to landuse activities, indicating that a 1 unit increase in vineyard cover and road density are associated with 20 to 25 percent declines in fish survival. In contrast, riparian vegetation cover had a significant positive effect on recruitment and survival. As with the landuse variables, riparian cover may have captured other unobserved yet important habitat variables, such as water temperature.

Journal Articles on this Report : 4 Displayed | Download in RIS Format

Other project views: All 10 publications 5 publications in selected types All 5 journal articles
Type Citation Project Document Sources
Journal Article Beche LA, McElravy EP, Resh VH. Long-term seasonal variation in the biological traits of benthic-macroinvertebrates in two Mediterranean-climate streams in California, U.S.A. Freshwater Biology 2006;51(1):56-75. R832439 (2006)
R832439 (2008)
R832439 (Final)
  • Abstract: Wiley Online
  • Journal Article Beche LA, Resh VH. Short-term climatic trends affect the temporal variability of macroinvertebrates in California 'Mediterranean' streams. Freshwater Biology 2007;52(12):2317-2339. R832439 (2007)
    R832439 (2008)
    R832439 (Final)
  • Abstract: Wiley Online-Abstract
  • Journal Article Beche LA, Resh VH. Biological traits of benthic macroinvertebrates in California mediterranean-climate streams:long-term annual variability and trait diversity patterns. Fundamental and Applied Limnology/Archiv fur Hydrobiologie 2007;169(1):1-23. R832439 (2007)
    R832439 (2008)
    R832439 (Final)
  • Abstract: IngentaConnect-Abstract
  • Other:
  • Journal Article Deitch MJ, Kondolf GM, Merenlender AM. Hydrologic impacts of small-scale instream diversions for frost and heat protection in the California wine country. River Research and Applications 2009;25(2):118-134. R832439 (2007)
    R832439 (2008)
    R832439 (Final)
  • Full-text: University of California-PDF
  • Abstract: Wiley Online-Abstract
  • Supplemental Keywords:

    RFA, Scientific Discipline, Ecosystem Protection/Environmental Exposure & Risk, Aquatic Ecosystems & Estuarine Research, Aquatic Ecosystem, Environmental Monitoring, Ecological Risk Assessment, Ecology and Ecosystems, anthropogenic stress, estuarine research, species interaction, ecological thresholds, salmon recovery, anthropogenic impact, ecosystem indicators, modeling ecosystem change, stream habitat, aquatic ecosystems, water quality, ecosystem stress, riverine ecosystems, trophic interactions, aquatic ecosystem restoration, ecosystem response

    Progress and Final Reports:

    Original Abstract
  • 2007 Progress Report
  • Final Report