Response:

Are the streams dry or just stagnant?

If dry, this is a regulatory issue. It reminds me of the problem where lakes have been drained and are now corn fields.

If stagnant, I would think they should be considered as lotic and would meet reservoir or lake criteria. I'm thinking of something such as a canal. In these cases, the floating algae rather than benthics should dominate so a water-based chlorophyll and nutrient criterion could be used. It may be that a DO criterion could be important.

Review:

If there is never flow, then nutrient criteria are a moot point. If discharge is intermittent or ephemeral there may be good reasons for setting nutrient criteria.

It only takes a week or so for algae to recover when a dry stream starts flowing again (Dodds et al. 1996). Similar recovery times are seen in desert streams after flooding (Fisher et al. 1982). It is fairly clear that recovery (growth) rates of attached algae in streams are related to nutrient concentration (Biggs 2000).

In situations where low flow occurs with diversion there may be very strong reasons for setting nutrient criteria. If algal biomass is high because nutrients are abundant, flow stagnates, then algal bloom problems concentrate and the chance for anoxia goes up. Isolated pools in such situations can be the last refuge for fish and a large amount of decomposing algal materials in these pools could lower dissolved oxygen. Since this is also a time of increased temperatures in many cases, the problem for fish would be exacerbated.

The Murray River in Australia is an example of a river where diversions and low discharge can lead to bad algal blooms (Burns and Walker 2000, Maier et al. 2001), and even toxic algal blooms (Baker and Humpage 1994). Cyanobacterial bloom problems in this river indicate the nutrient enrichment effect.

References
Baker, P. D. and A. R. Humpage. 1994. Toxicity associated with commonly occurring cyanobacteria in surface waters of the Murray-Darling Basin, Australia. Australian Journal of Freshwater Research 45:773-786.

Biggs, B. J. F. 2000. Eutrophication of streams and rivers: dissolved nutrient-chlorophyll. Journal of the North American Benthological Society 19:17-31.

Burns, A. and K. F. Walker 2000. Effects of water level regulation on algal biofilms in the River Murray, South Austrailia . Regulated Rivers: Research & Management 16: 433-444.

Dodds, W. K. R. E. Hutson A. C. Eichem M. A. Evans D. A. Gudder K. M. Fritz and L. Gray. 1996. The relationship of floods, drying, flow and light to primary production and producer biomass in a prairie stream. Hydrobiologia 333:151-159.

Fisher, S. G., L. J. Gray, N. B. Grimm and D. E. Busch . 1982. Temporal succession in a desert stream ecosystem following flash flooding. Ecological Monographs 52:93-110.

Maier, H. R., M. D. Burch and M. Bormans 2001. Flow management strategies to control blooms of the cyanobacterium, Anabaena circinalis, in the River Murray at Morgan , South Australia . Regulated Rivers: Research & Management 17: 637-650.

Review:

I concur with the first reviewer's opinion. It is important to establish whether there is flow and what the volume remaining is. If this is a riverine system made more or less permanently stagnant or non-flowing by diversion, then there are bigger problems for the system and perhaps a flow criterion is necessary to start. But, if these systems have been made intermittent or seasonal in terms of flow, then a nutrient criteria could be developed, albeit as a new class of stream, I would argue. Developing nutrient criteria would then develop along the lines of any system - looking at least disturbed reference sites to establish a reference database, considering nutrient-algal response relationships, perhaps some experiments to look at threshold concentrations, expert regional panel discussions, etc. In either case, there will likely be very little tolerance for nutrient inputs. These are essentially shallow (therefore not likely light limited), lentic systems - like small, shallow ponds. Their resistance to any nutrient input without excess plant growth will be minimal.

Response:

This question has 2 parts:  the first is the definition of impairment, and the second is the detection of impairment.  In terms of the Clean Water Act and water quality management, waters are impaired when they fail to meet water quality criteria for their designated use.  The Nutrient Criteria Technical Guidance Manual: Rivers and Streams (EPA 2000; EPA-822-B-00-002) has an excellent and concise discussion of water quality standards and criteria in its Introduction (pp. 1-15).

Excess nutrients can cause undesirable effects in several aspects of water quality (potential designated uses affected in parentheses): low dissolved oxygen (fisheries, recreation, aquatic life), turbidity (aesthetics, recreation, aquatic life), taste (drinking water, aesthetics), odor (aesthetics, contact), cyanobacterial toxins (drinking water, fisheries, contact, aquatic life), probability of pathogens (drinking water, contact), abundance and health of fish (fisheries, recreation, aquatic life), effects on condition of biota (aquatic life), etc.

The first step is to define what is being protected.  For aquatic life use, this is typically stated as “aquatic life as naturally occurs”.  In order to link nutrient criteria to aquatic life use, it will be necessary for a state to develop and implement biological criteria, preferably numeric criteria based on one of the many biological indexes now in use (see EPA 2005 (draft), Use of Biological Information to Better Define Aquatic Life Uses in State and Tribal Water Quality Standards: Tiered Aquatic Life Uses, US EPA, Office of Science and Technology).  A note of caution:  benthic macroinvertebrates are not the only assemblage for biocriteria – a second assemblage, say periphyton, or fish, is highly recommended to have a program sufficient to develop meaningful numeric biocriteria (Barbour and Yoder 2006 (draft), Critical Technical Elements of a Bioassessment Program, US EPA, Office of Science and Technology).  Periphyton and phytoplankton are recommended as the first assemblages of the biota to respond to changes in nutrient loadings.

Having biocriteria, how do we determine nutrient levels (i.e., nutrient criteria) that protect biological condition at the criteria level or better?  This is the difficult question, and is one of the primary reasons that the distribution approach has been to estimate reference nutrient levels, and develop criteria accordingly.  The the distribution approach avoids the research question altogether, with the quite reasonable assumption that nutrient levels and loadings similar to natural reference will not result in eutrophication.  For development of criteria based on response, periphyton and phytoplankton species composition are the assemblages that change the most predictably with nutrients.

Let us suppose that the work has been done and stressor-response data have been collected on biological response to nutrients.  For simplicity, assume a single biocriterion, not tiered aquatic life uses (see EPA 2005).  Consider the hypothetical figure below:  a biological response index shows a linear response to a nutrient index, given by the diagonal regression line.  All units are deleted from the figure to prevent any misinterpretation; it is to illustrate how criteria may be set.  The heavy horizontal line is the state’s biocriteria threshold value:  the biological condition must not fall below the value given by the line (13 in this case).  Water bodies below the line are considered impaired, and those at or above the line are considered unimpaired.

The (red) solid diagonal line is the regression line of the response to nutrients, and the 2 dashed diagonals represent (approximately) the 90% prediction interval for biological response given a nutrient concentration.  The vertical dotted lines show the nutrient values where the regression line and the 90% prediction intervals cross the biocriteria threshold.  These are 3 potential values for setting nutrient criteria.  For the first case, consider setting nutrient criteria at the point where the regression line crosses the biocriteria threshold, dotted vertical “b”.  If the nutrient concentration exceeds b, then action must take place (TMDL, permit limits, etc.).  The state can not regulate until and unless the nutrient value exceeds point b.  If all waterbodies are well and effectively regulated by the state, and all have point and non-point sources on them, then in the limiting case, all waterbodies will have a nutrient value corresponding to, but not exceeding, b.  

What are the predicted consequences if all waterbodies are at, but do not exceed, the nutrient criteria b?  The regression predicts that at nutrient concentration b, the average biological response of all water bodies will be at the biocriteria threshold.  This means that approximately half of all waterbodies will be impaired and half will be unimpaired (if the distribution of bio scores is symmetrical).  Thus, at the exact nutrient criteria concentration, the target protection level is 50% of waterbodies.  Protection of 90% of waterbodies would require nutrient criteria to be set at a.

If the primary objective of the state is to avoid unnecessary nutrient reduction, then the criteria should be set at point c, which results in only a 10% chance that a waterbody will be protected, but approximately 90% of nutrient reductions will have been necessary.  However, if the nutrient reductions did not go substantially below c, then most will have been insufficient.

The above approach implies that the nutrient criteria are not-to-exceed limits (or some allowance of no more than a certain percent for a certain fraction of time, e.g., 10% excursions for no more than 10% of time).  This is consistent with some toxic criteria, which are intended to protect, say, 95% of species all of the time (EPA 2002).

The part of the question on Type I and Type II errors implies the desire to make decisions based on statistical hypothesis tests, e.g., does this waterbody exceed the criteria with 95% confidence?  This is an hypothesis test where the null hypothesis is “H0: does not exceed”, and “Ha: exceeds criteria.”  The hypothesis test approach stands EPA’s standard toxic criteria approach on its head.  Instead of criteria being protective, and state regulation ensuring that harmful substances remain below the threshold criteria level, now the state will only act after it has been conclusively demonstrated that the substance has reached a harmful level.  Is this protective of water resources?

In order to remain conservative and protective of water resources, and use hypothesis testing, then the appropriate tests would be equivalence or noninferiority tests (e.g., Shukla et al. 2000, Streiner 2003, Piaggio et al. 2006).  For water quality criteria, a noninferiority test assumes as the null hypothesis that the criteria are exceeded by some preset, but small margin, and the alternative hypothesis is that the value is less than or equal to the criteria.  In order for the state to not impose management, it must be statistically demonstrated that the substance does not exceed the criteria.  To restate this approach, a waterbody is on the 303(d) list unless it is conclusively demonstrated to meet water quality criteria.  This provides an incentive for dischargers and other regulated entities to provide quality data.

Further thought and consideration on the issues discussed above will reveal the difficulties and management costs in imposing one-size-fits-all nutrient criteria to match one-size-fits-all biocriteria.  A tiered aquatic life use approach, which uses multiple biocriteria (EPA 2005), will result in more rational management of aquatic resources, including both better protection of outstanding resources, and better avoidance of unnecessary regulation.

References

Barbour, M.T., and C. O. Yoder. 2006 (draft). Critical Technical Elements of a Bioassessment Program. US EPA, Office of Science and Technology.

Piaggio, G., and others 2006.  Reporting of Noninferiority and Equivalence Randomized Trials.  An Extension of the CONSORT Statement.  J. Amer. Med. Assoc. 295:1152-1161

Shukla, R., Q. Wang, F. Fulk, C. Deng, and D. Denton.  2000.  Bioequivalence approach for whole effluent toxicity testing.  Environ. Toxicol. Chem. 19: 169-174.

Streiner, D.L.  2003.  Unicorns do exist: A tutorial on proving the null hypothesis.  Can. J Psychiatry 2003;48:756–761)

U.S. EPA.  2002.  Consolidated Assessment and Listing Methodology.  Towards a compendium of best practices.  First Edition.  US EPA Office of Wetlands, Oceans, and Watersheds.  http://www.epa.gov/owow/monitoring/calm.html

U.S. EPA. 2000.  Nutrient Criteria Technical Guidance Manual: Rivers and Streams  EPA-822-B-00-002

U.S. EPA 2005 (draft), Use of Biological Information to Better Define Aquatic Life Uses in State and Tribal Water Quality Standards: Tiered Aquatic Life Uses.US EPA, Office of Science and Technology.

Review:

I really like the response, both the definition of impairment and explanation of response.

Review:

The respondent has offered a pretty thorough response and I only make the following additions and support.

Impairment is defined as exceedance of a criterion. This definition is not particularly controversial. The better question, and perhaps the question you are getting at, is what variables to develop criteria for. EPA recommends developing parameters for response variables and nutrients. Response variables (chlorophyll and clarity) can more clearly be linked to use impairments, especially recreation and drinking water uses. Nutrients are not directly linked to use impairment, with some exceptions (ammonia and nitrate toxicity), but are, however, clearly linked to a variety of different responses that directly impact uses (algal toxin production, pH and oxygen impacts, impacts to aquatic community structure, etc). Nutrients, however, with some exceptions are both easier to measure and interpret than many of these other variables and can therefore, more adequately represent existing and potential risk, especially to downstream uses. Think of it like blood pressure. The pressure of blood per se may not harm a person directly; rather the resultant responses of the body to increased blood pressure are the problem (damage to blood vessels, etc.). However, these responses are so diverse and difficult and complicated to measure that it is easier and more efficient to use blood pressure as a measure of health risks to the body. So, too, nutrients are an efficient measure of the risk to a waterbody of developing the variety of responses that impact uses.

The definition of impairment does not depend on the method or variable selected; it is defined as exceeding a criterion. Maybe a more deliberative question is will the choice of the response variable affect the nutrient concentration that may cause use impairment? Perhaps, given the many variables that respond to nutrients and can impact uses. It is unlikely all of them will respond identically to nutrients and be related to use impairment identically. This argues, again, for the use of N and P criteria, since all of the many risks to uses can be assessed using nutrient concentrations alone.

The question of type I and type II error risk has been very adequately addressed in the response. I would reiterate that type II error, the error of not making an impairment decision when it truly exists, represents the greatest risk to human and environmental health. Too often we focus on type I error, more as a function of traditional statistical training rather than from truly thinking about risk I think, and set a conservative value for alpha, only to have beta remain high and power low, resulting in a significant risk of Type II error. Type I error is only a risk to dischargers of pollutants. Type II error is a risk to everyone and the environment. Refocusing on the power and type II error would likely encourage more rigorous testing and reporting.

Response:

There is some information on super-saturation of dissolved oxygen alone being harmful.

High concentrations of dissolved oxygen can inhibit photosynthesis of stream organisms

Dodds, W. K. 1989. Photosynthesis of two morphologies of Nostoc parmelioides (Cyanobacteria) as related to current velocities and diffusion patterns. J. Phycol. 25:258 262.

Hyperoxic/free radical damage to organisms is increased with increasing dissolved oxygen, so there may be sensitive species in all groups of animals.
D. Jamieson, B. Chance, E. Cadenas, and A. Boveris. The Relation of Free Radical Production to Hyperoxia. Annual Review of Physiology. Vol. 48: 703-719

However, there are not many papers that I am aware of on this effect on animals in streams specifically.

Review:

The only supersaturation literature I have seen, such as that cited in the EPA 1986 Quality Criteria for Water (Gold Book)(attached to answer)(EPA440/5-86-01), indicates that it is total dissolved gas supersaturation which leads to physiological problems in fish and that, while supersaturation effects have been noticed in response to oxygen, it is principally nitrogen supersaturation that is critical. The recommended criterion is 110% supersaturation - of total dissolved gases. Remember, too, that oxygen is about 35% of the dissolved gas in water, the rest being nitrogen and oxygen is bioactive whereas nitrogen is essentially inert to most organisms. So, you can have and likely often do have oxygen supersaturation without any effects since nitrogen is the lion's share of dissolved gases, is more lethal, and its saturation may offset oxygen supersaturation if it is undersaturated. TDS is difficult to measure, as well, so it is hard to tell from a simple DO saturation measure whether you have a supersaturated condition. Much of the effects literature cited in the 1986 guidance is from salmonid research (inverts apparently react at higher supersaturation levels) and much of the supersaturation of concern in those studies was from discharges and/or dam spillage. So, to cut to the chase, oxygen supersaturation will likely only cause problems when it contributes to a total dissolved gas supersaturation greater than 110%.

We have not seen much on the effects of large swings. Nothing that suggests that the swing magnitude per se rather than the minimum concentration is what matters. The research we have seen suggests that swing magnitude above DO minima does not really matter - what matters is falling below the minima and the duration of time spent below the minima.

Review:

High photosynthetic activity by phytoplankon in lakes, or periphyton in streams, can have 2 direct chemical effects: supersaturation of dissolved oxygen, and depletion of carbonate and CO2 from uptake. Since dissolved CO2 acidifies water by the formation of carbonic acid, removal of CO2 raises the pH. Both of these effects are strongly diurnal, so that DO and pH fall during the night.

Gas supersaturation can cause "gas bubble disease" or "gas bubble trauma", lesions caused by the gas coming out of solution and forming bubbles in tissues - essentially the same thing as "the Bends". It can occur in vertebrates and invertebrates. In fish it is most commonly associated with hydropower dam tailwaters (gas in equilibrium in the hypolimnion is supersaturated at surface pressures, as well as aeration over spillways), thermal effluents in winter (cold water in equilibrium becomes supersaturated when heated), and excessive aeration in hatcheries and ponds. The bubbles are composed of the constituent gases, and since N2 is nonreactive and the dominant gas (79% in air, 65% in water), most damage is caused by nitrogen.

A small number of papers have attributed gas bubble trauma to excess oxygen:
A bloom of Chlamydomonas caused DO as high as 30-32 ppm (> 300% saturation) and was associated with a fish kill, where the dead fish had characteristic gill and skin lesions of gas bubble disease (Woodbury 1942).
A similar situation ocurred in Galveston Bay, Texas, with fish mortality observed coincident with DO of 250%, following an algal bloom (Renfro 1963). Weitkamp and Katz (1980) cite these and several other papers that documented the same phenomenon, for both fish and invertebrates.

Oxygen supersaturation may add to multiple stressors without in itself being sufficient for mortality. Mortality of trout with whirling disease increased when the fish were subjected to additional stressors, including supersaturation (Schisler et al. 2000); some species of abalone have reduced growth rates when grown in oxygen supersaturation; and and some showed increased incidence of Vibrio infection compared to those grown in normal conditions (Elston 1983, Harris et al. 2005). Finally, translucent fish larvae may ingest abundant phytoplankton in a bloom, which continue to photosynthesize and produce oxygen in the larval guts (Mishra and Kumar 1997) - although not a direct effect of supersaturation.

In highly productive streams and lakes in summer, diurnal pH fluctuation may be as much as 1.5-2 pH units, although fluctuations less than 1 unit are more common (Krstolic and Hayes 2004). Extreme fluctuations require high levels of enrichment, and these water bodies tend to be relatively alkaline, hard water systems with "normal" pH in the range 7-8.5. pH values above 9.0 are thought to be harmful to some aquatic organisms, either from direct toxicity of hydroxide ions, or mobilization of some toxic metals. In addition, high pH (and warm temperature) drives ammonium ions to unionized ammonia, so it is possible to have episodic ammonia toxicity due to pH if total ammonia is high enough. Most of these effects require extreme algal growth, characteristic of hypereutrophic conditions. Mortality of Ceriodaphnia reticulata was attributed to pH fluctuation in small eutrophic ponds (O'Brien and DeNoyelles 1972).

Elston, R. 1983. Histopathology of oxygen intoxication in the juvenile red abalone, Haliotis rufescens Swainson. J. Fish Diseases 6:101-110.

Harris, J.O., C.M. Burke, S.J. Edwards, and D.R. Johns. 2005. Effects of oxygen supersaturation and temperature on juvenile greenlip, Haliotis laevigata Donovan, and blaklip, Haliotis rubra Leach, abalone. Aquaculture Res. 36:1400-1407.

Krstolic, J.L., and D.C. Hayes. 2004. Water quality synoptic sampling, July 1999: North Fork Shenandoah River, Virginia. U.S. Geological Survey, Scientific Investigations Report 2004-5153

Mishra, B.K., and D. Kumar. 1997. Observation of gas bubble disease in postlarvae of Indian carps important to aquaculture. J. Aquat Animal Health 4:106-108.

O'Brien, W.J., and F. DeNoyelles, Jr. 1972. Diurnally elevated pH as a factor in zooplankton mortality in nutrient enriched ponds. Ecology 53: 605-614

Renfro, W.C. 1963. Gas-bubble mortality of fishes in Galveston Bay, Texas. Trans. Am. Fish Soc. 92:320-322.

Schisler, G.J., E.P. Bergeresen, and P. G. Walker. 2000. Effects of multiple stressors on morbidity and mortality of fingerlin rainbow trout ing=fected with Myxobolus cerebralis. Trans. Am. Fish. Soc. 129: 859-865.

Weitkamp, D.E., and M. Katz. 1980. A review of dissolved gas supersaturation literature. Trans. Am. Fish. Soc. 109:659-702.

Woodbury. L.A. 1942. A sudden mortality of fishes acoompanying a supersaturation of oxygen in Lake Waubesa, Wisconsin. Trans. Am. Fish. Soc. 71: 112-117.

Context:

As we move forward with building our database of water quality paired with attached algae, etc., I want to make sure we have low enough detection limits to be of use in setting nutrient criteria, or at least screening levels. Currently, our lab's detection limit for nitrate+nitrite is 0.1 mg/l. When all of our data are examined, about 60% are below detection. This seems to tell me that a lower detection limit may be more useful.

Response:

Yes, those limits of detection are too high. In a typical oligotrophic system, nitrate should barely detectable or at least often below 10 ug/L. Ammonium and SRP are often below 1 ug/L (0.001 mg/L). Nutrient criteria should be based on total N and total P. Those values in oligotrophic systems often fall below 100 ug N/L and 10 ug P/L. For a reference on how these values fall out across the country for reference stream sites see:

Dodds, W. K.; Oakes, R. M. A technique for establishing reference nutrient concentrations across watersheds affected by humans. Limnol. Oceanogr. Methods 2004, 2, 333-341.

Smith, R. A.; Alexander, R. B.; Schwarz, G. E. Natural background concentrations of nutrients in streams and rivers of the conterminous United States. Environ. Sci. Technol. 2003, 37, 3039-3047.

Keep in mind that lake concentrations are generally less than those found in streams (because of settling).
For potential problems with using DIN and SRP to determine nutrient criteria see:

Dodds, W. K. 2003. The misuse of inorganic N and soluble reactive P to indicate nutrient status of surface waters. Journal of the North American Benthological Society 22:171-181.

Review:

I agree, the detection limits you are using are too high for nutrient criteria development. Once criteria are in place, detection (and quantitation) limits should be lower than the criteria so that you know when the criteria are exceeded. Since you are developing the criteria, you may want them lower still to make sure you can detect the effects of enrichment so that criteria can be developed. I also agree that total N and total P are the most relevant measures for nutrient criteria. The papers cited should give you an indication of detection limits that would be relevant for Nevada.

Review:

I agree with both reviewers. It has become a problem for many states to develop nutrient criteria due to high detection limits from the earlier year's data. I also agree that nutrient criteria should be built based on TN and TP. However, it is quite common that nitrate and nitrite concentrations are better measured (lower detection limits) than TN because of high detection limits of TKN in many labs.