STREAM TEMPERATURE INVESTIGATIONS: FIELD AND ANALYTIC METHODS INSTREAM FLOW INFORMATION PAPER: NO. 13 Biological Report 89(17) June 1989 AQUATIC SYSTEMS BRANCH TEMPERATUREERRATA FOR INFORMATION PAPER 13 Bartholow, J.M. 1989. Stream temperature investigations: field and analytic methods. instream Flow Information Paper No. 13. U.S. Fish Wildl. Serv. Biol. Rep. 89(17). 139 pp. The following error was noted after printing. We sincerely apologize for any inconvenience this may have caused you. Page 116, last reference Delete reference to Smith, K. 1975. See instead Smith, K., and M.E. Lavis (1975) as referenced on page 117.Biological Report 89(17) June 1989 STREAM TEMPERATURE INVESTIGATIONS: FIELD AND ANALYTIC METHOOS Instream Flow Information Paper No. 13 by John M. Bartholow U.S. Fish and Wildlife Service National Ecology Research Center 2627 Redwing Road Fort Collins, CO 80526-2899 U.S. Department of the Interior Fish and Wildlife Service Research and Development Washington, DC 20240DISCLAIMER The mention of trade names does not constitute endorsement or reconmenda- tion for use by the Federal Government. Suggested « itation: Bartholow, J.M. 1989. Stream temperature investigations: field and analytic methods. Instream Flow Information Paper No. 13. U.S. Fish Wildl. Serv. Biol. Rep. 89 (17). 139 pp.‘SUMMARY This document provides guidance to the user of the U.S. Fish and Wildlife Service’s Stream Network Temperature Model (SNTEMP). Planning a temperature study is discussed in terms of understanding the management objectives and ensuring that the questions will be accurately answered with the modeling approach being used. A sensitivity analysis of SNTEMP is presented to illustrate which input. variables are most important in predicting stream temperatures. This information helps prioritize data cotlection activities, highlights the need for quality control, focuses on which parameters can be estimated rather than measured, and offers a broader perspective on management options in terms of knowing where the biggest temperature response will be felt. All of the major input variables for stream geometry, meteorology, and hydrology are discussed in detail. Each variable is defined, with guidance given on how to measure it, what kind of equipment to use, where to obtain it from another agency, and how to calculate it if the data are in a form other than that required by SNTEMP. Examples are presented for the various forms in which water temperature, discharge, and meteorological data are commonly found. Ranges of values for certain input variables that are difficult to measure or estimate are given. Particular attention is given to those variables not commonly understood by field biologists likely to be involved in a stream temperature study. Pertinent literature is cited for each variable, with paphasis on how other people have treated particular problems and on results they ave found. Model calibration, verification, and validation steps are defined and outlined, with measures of "goodness-of-fit" given for comparing simulated stream temperatures with observed values. The question of how good is good enough is explored, and attention is given to the kinds of simulation and data reduction errors that one should be alert for. Some special cases dealing with ice and reservoir temperature are mentioned. Special attention is given to understanding micro-thermal habitats that act as important thermal refugia under low flow conditions; their causes, extent, and management implications are discussed. Alternative public domain stream and reservoir temperature models are contrasted with SNTEMP. A distinction is made between steady-flow and dynamic-flow models and their respective capabilities. Regression models are offered as an alternative approach for some situations, with appropriate mathematical formulations suggested. iiiAppendices provide information on State and Federal agencies that are good data sources, vendors for field instrumentation, and small computer programs useful in data reduction.CONTENTS SUMMARY .. FIGURES . TABLES .... ACKNOWLEDGMENTS . INTRODUCTION PLANNING A TEMPERATURE STUDY 5 UNDERSTANDING WATER TEMPERATURE THROUGH SENSITIVITY ANALYSIS . 6 DATA GATHERING AND FIELD TECHNIQUES .. 12 Stream Geometry Components 12 Meteorological Components 34 Hydrology Components ... 51 Water Temperature Components 54 Special Cases ... 74 SIMULATION AND ANALYSIS TECHNIQUES . 85 Alternative Predictive Techniques 85 Calibration/Validation ......... 89 Other Stream Temperature Models 96 Special Cases 101 LITERATURE CITED 1.0.2... sec eeece cece ene e eee s teense eee scence eenee eee aees 107 APPENDICES A. Data Sources .... + 120 B. Equipment Vendors . 121 €. Program Snippets .. 123 0, Alternative Data Formats from the Hydrodata Data Base 127 E. Selected Meteorological Data for July 132Number oe ww omouw an FIGURES Approximate "temperature-1imited" geographic range of rainbow trout under three scenarios ..... tote eeeeeseee teete eset cece ees . Various ways in which heat either enters or leaves a flowing SEVOAM vee eee seceeeeeecsene ee eeeeeee preteens pete eres . Sensitivity of the SNTEMP model’s predictions of mean daily water temperature to changes in various input parameters ...... Sensitivity of the SNTEMP model’s predictions of maximum daily water temperature to changes in various input parameters .. Development of width versus flow relationships showing no effect of units of measurement. on slope, but definite effect on intercept ............ tenes Variation in travel time versus d: along a stream ..... Schematic of dye conceit during time-of-travel study .........cccce0e0e eres Example of diel change in water temperature through time following clearcutting .............. Se eteeeeeees see eeeeeaeeeee Effect of one timber management alternative on hourly water temperatures .. deeeeee Pee e eee eeeeeee Schematic showing the interception of solar radiation by topography and riparian vegetation ......... tte eeeeerenee bene Conventions for determining the sign and degree of stream orientation as well as east and west bank designations . The “Bovee" home-made clinometer using a protractor, fishing weight, and soda straw ......... Riparian vegetation shade parametel Influence of buffer strip width on shade densi yo. we Frequency distribution of temperature lapse rates at different altitudes ..........00.. Petes ee eeeeeee eeeeee Field-installed meteorological station ...... sane Variation of amount of usable data from self-installed meteorological stations ............. kta e betes eeaeenenene Variation of air temperature above the Caney Fork River, TN Variation of relative humidity above the Caney Fork River, TN Total hemispheric radiation taken at Fort Collins, CO, during December 1988 fete eee teeter renee sone Effect of shelter belt on various wind speed classe: Effect of various shelterbelt heights on wind speed Annual variation of ground temperatures at Fort Collins, CO .. Departure from the mean ground temperature according to direction of slope and time of year ............. seeeeesFIGURES (Concluded) In situ and remotely configured thermographs ....... Priorities for installation and collection of water ure monitoring locations for various stream network configurations 66 Cross-sectional temperatures showing degree of temperature variation for a given transect at a given flow ...............5 68 The different types of error that can arise in the measurement of temperature ..........eeeee eee eee eee 72 Temperature versus distance versus time “display” . Common areas in which to expect temperature-created microhabitat 80 Temperature profiles within a pool at various flows ............. 82 Comparison of water temperatures in pools and upstream riffles .. 83 Full "split set” calibration/validation scheme ...........65 91 A spectrum of perspectives on possible causes of model error” 94 Duration plot for 30 years of daily air temperature from June 15 to August 15 . setetereeeneseseeeesees 104 Duration plot for 30 years of mean daily air temperature from dune 15 to August 15... veseeeee 104 Normal daily average temperature for duly 132 Mean relative humidity for July ....... 133 Mean percentage of possible sun for July 133 Mean daily solar radiation for July . 134 Mean wind speed for July ..........06 134 viiCe OD NM Hm ew nw D-1 0-2 0-3 D-4 TABLES Relative sensitivity of maximum and mean water temperatures to various parameters for a generalized stream ... Range of input values used to determine relative Of SNTEMP ...... 0 eee ceases Percentage of incident solar radiation in water... ... cece eee Availability of useful data e! fi Climatological Data Summaries tbe estes eee eeetesseneeeseeenen Percent reflectivity for various homogeneous ground cover conditions ........ Pier ttt n ects eeenteeenenees oe Day length for various time frames and latitudes ...... Alternative forms in which you may find water temperature data ious depths in USGS publications or data bases ............4.... seteeeeee One form in which you may find water temperature data from a STORET data base .............. Example thermograph specifications eee oe Average size, depth, and temperature depressions o} water areas found by Bilby in a Washington stream ............. Textural output file from Hydrodata for mean daily water temperature ............. Textual output file from temperature .... Part of a Lotus r maxinum daily air” eet created by Hydrodata Optica Disk Database 0.0... eee cececesceessususaseeee Textual output file from Hydrodata for streamflow viii 131ACKNOWLEDGMENTS: Many people are due credit for this paper. Kay Lindgren and her staff spent many hours finding and obtaining documents for me. Terry Waddle reviewed an early draft and helped to firm up the contents. Other reviewers included Carl Armour, De?] Simmons, Dave Gilbert, Fred Theurer, and Ken Voos. Special thanks are due to both John Homa and Wayne Lifton for extra- detailed review and comments. Kenny Dinan and Bill Slauson served to bring a real-world tone to the manuscript. Jenny Shoemaker and Mary Sanz did a fine Job of making or obtaining the figures, and Trish Gillis made the manuscript presentable. Jim Zuboy provided a much-needed editorial review. Finally, thanks to Bob Milhous for pressuring me to do this job in the first place. ixINTRODUCTION Water temperature has always been considered one of the most important factors determining the geographic distribution of fish and other aquatic organisms. Analysis of water temperature regimes has lately taken on added importance, primarily for economic reasons. A recent newspaper article (Rocky Mountain News 1988) discusses the construction of a $5.5 million reinforced- plastic curtain in northern California’s Shasta Dam to transfer cool water from the reservoir into the Sacramento River to prevent salmon heat death. Without this curtain, the Bureau of Reclamation must release water without passing it through turbines at a cost of $70,000 per day in lost power revenue. Another study (Croley et al. 1981) has shown that the incremental cost of reducing thermal discharges to achieve a 3 ‘F {1.7 °C) reduction along the Missouri and Mississippi Rivers would be about $211 million per year. This study measured power losses, but did not attempt to quantify fish and wildlife gains. In contrast, a paper by Theurer et al. (1985) developed a valuation of $0.6 million per year for restoring a salmon population in the temperature degraded Tucannon River, a small Washington State river with a present worth of $6.9 million. Clearly, these are large numbers, no matter which perspective one chooses. Recent climatic changes have caused the earth’s surface to be warmed by about 0.5 °C (0.9 °F) between 1861 and 1984, along with a decrease in the diurnal temperature range of about 1 °C (1.8 °F) (Zoltai 1988). A future perspective on water temperatures is even more interesting. The American Association for the Advancement of Science (1988) has concluded that scientists do not agree whether global warming is a reality or not. But it feels that it is at least prudent to look forward to what changes may be expected if the earth’s atmosphere were to warm by 1.5-4.5 °C (3-8 °F) (Smagorinsky 1982). In addition to the flooding of coastal cities and other alarming large-scale problems, we might expect both a lesser amount of precipitation and higher air {and water) temperatures, especially in the western United States. If this were to occur, species now at the margins of their thermally defined geographic range may be expected to change rather dramatically (Figure 1). The purposes of this report are many. First, it is intended to serve as a companion to another report in this series, Instream Flow Information Paper No. 16--Instream Water Temperature Model (Theurer et al. 1984), in which the theory and application of the Stream Network Temperature Model (and the SNTEMP set of computer programs) are described. The information presented here will serve to broaden some of the concepts and methodologies outlined in that publication, especially in the area of field techniques and laboratory analytical methods. Second, information and advice on other data collection procedures willCURRENT "TEMPERATURE-LIMITED” TROUT HABITAT Figure la. Approximate "temperature-1limited" geographic range of rainbow trout (Oncorhynchus iss, formerly Salmo gairdneri) under three scenarios: current conditions. Range defined by upper limit of 23.8 °C (75 °F) mean monthly surface water temperature. Figures derived from Hydroscience (1971).“TEMPERATURE-LIMITED” TROUT HABITAT IF GLOBAL WARMING OF 2.7°C Figure 1b. Approximate “temperature-limited" geograp! scenarios: assumine hic range of rainbow trout under three g a 2.7 °C (5 *F) uniform global warming.“TEMPERATURE-LIMITED” TROUT HABITAT IF GLOBAL WARMING OF 5.6°C Figure lc, Approximate "temperature-limited” geographic range of rainbow trout under three scenarios: assuming a 5.5 °C (10 °F} uniform global warming.be presented. Numerous questions always arise as to where and how inputs to any temperature model can or should be obtained. For each model input, I explain what it is, what’s known about it, and how to measure or estimate it. Third, ideas on what constitutes proper calibration/validation for temperature models are discussed, as there seems to be a lot of confusion over these and similar terms. I discuss what the terms mean, how to do things the "right" way, and when to do what. Finally, 1 give a brief review of alternative temperature models that may be used in place of the SNTEMP set of models, and a brief review of reservoir and other water quality models that may be used in conjunction with stream temperature models. One reviewer noted that this report will not be a "bestseller"; its audience is fairly specific, though the material is broad and diverse. It is directed towards those who have at least a general knowledge of Theurer’s Stream Network Temperature Model and want to become more proficient in planning field activities or engaging in simulation/analysis techniques. Others may benefit from information contained here, but that is not the primary purpose. PLANNING A TEMPERATURE STUDY The Aquatic Branch is constantly reminded by users of our models, especially the more complicated SNTEMP-type models, that we need to stress the need for careful study design. We assume that this means that there js difficulty in making sure that you (1} are going to be answering the right questions, (2) are using the right set of tools, and (3) can trust your answers. The Aquatic Branch has written volumes on laying out a study plan (Bartholow and Waddle 1986) and scoping questions to ask (Bovee 1982). These publications have not seemed to dent the continued insistence that study design is critical. Therefore, we can only conclude that adequate study plans are not being assembled. Some have suggested that better prestudy involvement between all_members of the “team" needs to be stressed. That is, planners, field data collectors, modelers, statisticians, decision makers, regulators, resource interests, developmental interests, and reviewers all need ACTIVE involvement to (after Henriksen, 1988): (1) identify the management problem (goals and objectives). Does this study deal with water rights or flow reservations? Is it to assess project impacts, evaluate mitigation, or approve permits? If it is an impact analysis problem, what is the appropriate baseline period with which to compare impacts? Are we at the feasibility or operational stage in the planning process? Is this a single project or a network of projects? Who are the players; who has the lead? How "important" is this project; is there a lot of resistance to a study of this type? (2) identify the appropriate species/life stages of concern. Is this a game, sport, or commercial fishery problem? Is it a sensitive or indicator species problem? Is it an endangered species problem? 5Is it a "guild" of species or a planned introduction? Are we talking about a naturally sustaining population, supplemental stocking, or @ put and take fishery? Do we have adequate life history information for periodicity, microhabitat preferences, and water quality? (3) identify the relevant variables to be measured/predicted. Is minimum, mean, maximum temperatures, or some combination the issue? Is a daily, weekly, or monthly averaging period appropriate? What is the spatial extent of your study area? (4) identify the appropriate criteria to employ. Are we talking about growth, mortality, trigger temperatures, temperature change rates, “minimum” flows, available fish habitat, population size, dollars, or commercial or recreational fishing effect? Do not proceed until criteria have been formulated and agreed to by all parties. (5) identify the quantitative measures for decision making (miles of suitable stream, temperature-conditioned microhabitat, hatching times, etc.). How concerned must we be about accuracy and/or their jens Do different players need different information to do their job? (6) identify and evaluate the feasible solution methods. Is adequate information already available to make the decisions at hand? If not, what techniques will best address the questions? Is there a favored method which has been used by local agencies? How much time, money, and manpower can (or should) be devoted to the problem and solution analysis? What is the time frame for decisions to be made? Can field studies be scheduled? What are realistic management options? We hope that by getting al} of the participants to reach a consensus on the above points, you will have come a Tong way toward resolving the impediments. that sneak up on otherwise well planned and executed studies. The remainder of this document is devoted to helping you perform at least the temperature analysis effectively and efficiently. UNDERSTANDING WATER TEMPERATURE THROUGH SENSITIVITY ANALYSIS Prior to any extensive water temperature modeling or analysis activity, it is wise to understand the influences that various stream geometry, meteoro- Jogical, and hydrological components have on determining water temperature. Such an understanding will better enable you to (1) prioritize data collection activities, (2) know the degree to which you should be concerned with quality contro) errors, (3) know which parameters can be safely estimated, and (4) broaden your perspective of potential management strategies. To further this understanding, we propose an initial consultation with a sensitivity analysis tool.Sensitivity analysis of deterministic models is a valuable step in any model application. There are several specific uses for sensitivity analysis, some for the model builder and tester, and some for the model practitioner. Sensitivity analysis may be used to {1) serve as an aid in confirming that the mode] is consistent with theory, (2) show the effect that errors in each parameter have on the dependent variable (water temperature), (3) identify those parameters that are sensitive to the degree that they warrant very reliable measurement, and (4) show the relationship between the parameters subject to management contro] and the dependent variable (Reckhow and Chapra 1983). For our purposes, it will be valuable to learn where to concentrate data collection efforts, and how to display the effect that changes in flow, riparian shade, or channel characteristics have on stream temperature. There are many ways of performing a sensitivity analysis on deterministic models. A common approach is a test in which a single parameter is systemati- cally varied, while other parameters are held constant, and the response of the dependent variable is monitored. This allows us to say, for example, "A unit change in X produces a 2% change in stream temperature.” A disadvantage of this technique is that it does not allow the practitioner to say what portion of the variance is attributable to a single parameter if the other parameters are also changing. Table 1 illustrates the relative sensitivity of the key parameters used as input to most temperature models. This analysis was performed for the SNTEMP model, many major components of which are illustrated in Figure 2, but the results would be expected to be similar across other deterministic stream temperature models. This table was generated by systematically varying the input parameters and noting the conditions associated with maximum changes in both mean and maximum water temperatures. This method gives a more robust picture of true sensitivities than varying a single parameter for only one set of other variables. It does not, however, explicitly consider the cross-correlation between parameters. The parameters in Table 1 are ordered down the page from most to least sensitive for the generalized stream being simulated. Other streams will behave differently, but the general pattern should remain relatively stable. There are some obvious exceptions, however, such as the case of water temperatures immediately downstream of a reservoir, where the primary influence on temperature is the release temperature itself. Parameters were varied for a generalized stream segment for an idealized July condition. The high and low values chosen to characterize this stream are shown in the Table 2. Items not shown in Table 2 were held constant; these values are: lateral flow, zero; upstream elevation, 100 feet; downstream elevation, zero feet; segment length, 10 miles; width’s B value, .2; day length, 14.5 hours; and dam at inflow, true. Clearly, a headwater stream or a large river’s parameters would be different.Table 1. Relative sensitivity of maximum and mean water temperatures to various parameters for a generalized stream. Sensitivity as depicted here is dimensionless. Please see text for an explana- tion. When these associated parameters are combined 2s show: To changes. in Stream Inflow Width/ Thermal Air Relative Wind Solar Percent Water temperature is: these variables: flow temp. depth gradient temp, humidity speed radiation shade Very Sensitive Air Temperature Tow low high high high Moderately Sensitive Percent. Shade low high Tow high Moderately Sensitive Relative Humidity Joe low high high low Hoder: Stream Flow Tow Tow high high Tow Moderately Sensitive Inflow Temperature high high low Yow Tow high Woderately Sensitive Stream Width Jom low high high Tow Relatively Insensitive Soler Radiat ion ow high low low Tow Relatively Insensitive Travel Time/Roughness Tow Tow high low Relatively Insensitive vind Speed high Tow high high high Relatively Insensitive Ground Temperature Tow high high Yow Insensitive Percent Possible Sun low high high Yow Tow Insens{t ive Thermal Gradient low high high highHEAT FLUX SOURCES lye N 2 i ATMOSPHERIC RADIATION \ Carnie 7 SOLAR saoiarion _/ bust SCATTERING - wt RIPARIAN VEGETATION RADIATION REFLECTION ALSO: (1) HEAT LOSS DUE TO EVAPORATION (2) HEAT GAIN DUE TO FLUID FRICTION (3) HEAT EXCHANGE DUE TO AIR GIRCULATION (CONVECTION) cs locaupep CONDUCTION Figure 2. Various ways in which heat either enters or leaves a flowing stream. Adapted from Theurer et al. (1984).Table 2. Range of input values used to determine relative sensitivity of SNTEMP. Parameter Low vatue High value Units Inflow 10.0 20.0 cfs Inflow temperature 6.0 12.0 “Cc Roughness +035 055 NA Width’s A 13.5 18.0 NA 2 Thermal gradient 1.5 1.8 d/m*/sec/*C Air temperature 65.0 85.0 °F Relative humidity 40.0 70.0 percent. Wind speed 6.0 9.0 mph Percent possible sun 60.0 80.0 percent. Solar radiation 495.0 630.0 Langleys Segment shade 25.0 75.0 percent Ground temperature 10.0 16.0 "c In Table 1, water temperature is very sensitive to changes in air temperature when stream flow is low, inflow temperature is low, width-to-depth ratio is high, relative humidity is high, and wind speed is high. Water temperature may be sensitive to air temperature when these conditions are not present, but it will not be as sensitive. Water temperature is insensitive to changes in thermal gradient all the time. However, changes in thermal gradient cause the most change in water temperature when stream flow is low, width-to-depth ratio is high, air temperature is high, and relative humidity is high. Note that the entry for travel time/roughness applies only to maximum water temperatures; it does not effect mean daily water temperatures in the SNTEMP model. Another way to look at the relative sensitivity of water temperature to changes in model variables is to plot the absolute change in predicted temperature produced by varying the parameters through the same combinations displayed above. The range of values so produced can be large. It is instructive to plot the data by quartiles, showing the minimum, maximum, and median values. Graphs for the mean and maximum water temperatures are shown in Figures 3 and 4. The values for mean and maximum are similar except for shade, solar radiation, and roughness.SENSITIVITY OF MEAN DAILY WATER TEMPERATURE, co 1 2 3 4 8 6 rt eg 8 ww Af TEMPERATURE, RELATIVE HUMIDITY. PERCENT SHADE STREAM FLOW (FLOW TEMPERATURE stmeAu wor oo LeaeNo ’ sous naoanion | ea oe we sete0 a LES ES erowe romerrine fe (E(t ooh PERCENT POGSIBLE SUN je ‘THERMAL GRADIENT ROUGHNESS ot @ 8 « 6 68 F Bo 49 RELATIVE CHANGE IN MEAN DAILY WATER TEMPERATURE Figure 3. Sensitivity of the SNTEMP model's predictions of mean daily water emperature to changes in various input parameters. SENSITIVITY OF MAXIMUM DAILY WATER TEMPERATURE Ca a ee AIR TEMPERATURE PERCENT SHADE RELATIVE HUMODITY STREAM FLOW MIFLOW TEMPERATURE STREAM wioTH SOLAR RADIATION “eee os sovomess Lee wn tren cde ‘GROUND TEMPERATURE PERCENT POSSIBLE SUN THERMAL GRADIENT eo te sll RELATIVE CHANGE IN MAXIMUM DAILY WATER TEMPERATURE Figure 4. Sensitivity of the SNTEMP model's predictions of maximum daily water temperature to changes in various input parameters. oatPrevious authors suggest that factors other than flow largely dictate water temperatures (Jowett and Mosley 1983; Laenen and Hansen 1985). Few authors have dealt, with the sensitivity of water temperature models to a variety of parameters! (Moore 1967; Crittenden 1978). Crittenden’s (1978) sensitivity analysis differs from my example and from other authors’ in several respects. First, he varied only a single parameter at a time, and second, the model he used was developed solely for predicting equilibrium temperatures in small, unshaded, Tow gradient streams with little groundwater inflow. His results indicate that wind speed and the thermal properties of the substrate are the two most sensitive parameters. I find these conclusions suspect because these two parameters were varied over two orders of magnitude in the case of wind and one order of magnitude for thermal diffusivity. I do not believe these are reasonable variations for "rea? world" applications. In summary, I strongly advise that a sensitivity analysis, even if crude, be performed prior to any field work or other data collection to determine which Parameters deserve special attention. Do not take the examples given here as necessarily indicative of your situation. DATA GATHERING AND FIELD TECHNIQUES Armed with a general knowledge of which Parameters are most likely controlling water temperature, we can proceed to the discussion of individual model parameters--what they mean and how to estimate them. Our discussion will be divided into three major groups: stream geometry, meteorology, and hydrology. STREAM GEOMETRY COMPONENTS Elevations, distances, and stream widths are fundamental stream geometry measurements, These get respectively more sensitive and also more difficult to calculate accurately. Elevations Elevations are important in temperature modeling for (1) calculating the slope resulting in heat from friction, (2) calculating the atmospheric pressure, an important element in heat convection, (3) calculating the depth of the atmosphere through which solar radiation passes, and (4) translating known air When reviewing their work, it is well to remember that there may not be an adequate distinction between air temperature and solar radiation, Accordingly, there may be confusion between proximate and ultimate causes in the sense that short wave solar radiation warms the air, which in turn emits long wave thermal radiation. In terms of heat flux, atmospheric radiation dominates most of the time, especially in the summer. Occasionally, the sensitivity discussion in the literature must be interpreted as Sensitivity of maximum daily water tempera- tures, not mean daily. wztemperatures and relative humidities to points of known elevation. Though any of these major processes may be of great importance, none are individually sensitive to small errors in elevation. Thus, elevations may be taken from readily available topographic maps even though contour intervals on some maps may be 40 feet and low relief terrain may not have easily discernable elevations. The most difficult task may be identifying where some station or node actually is on the map. For example, field work may have indicated significant changes in the distribution of riparian vegetation not apparent on the map. Be wary of trying to model very steep gradient, almost waterfall, situations; SNTEMP may dramatically overestimate heat flux due to friction if the stream width is too narrow. Distances Stream distances are important in calcutation of heat transport. Distances basically translate to travel time and thus exposure time to all of the heat flux conditions. Aside from river mile indices that may be available, maps or aerial photos of known scale provide the easiest way to estimate distances. Distances can be a source of model bias if consistently over- or underestimated. Streams, being sinuous, can be tricky to measure reliably using a map and a map wheel, especially if the map has been protected with an acetate cover. It is best to measure the segments repeatedly, using a paper map, and take an average. In cases where you are aware that a schematic map does not convey the true sinuosity, it may be advisable to multtply the measurement by a "fudge factor" to account for the difference. It would be better to use aeria} photos in this Situation, If a temperature analysis is being conducted in conjunction with a Physical Habitat Simulation Model (PHABSIM) analysis (Milhous et al. 1989), actual surveyed or paced stream segment distances may be available, especially if the detailed “habitat mapping” approach (Morhardt et al. 1983) is being used, It is always preferable to measure the distance the water is actually flowing. Also, distance may actually change as a function of flow. If large changes are anticipated, adjustments to model distances must be made in a fashion analogous to PHABSIM high-flow and low-flow models. That is, one set of data is used to describe the high flow conditions, one set the low flow. Stream Width Stream width can be a very sensitive parameter (recall Figures 3 and 4) in modeling water temperatures. All of the heat flux activities take place at either the air-water interface or the water-ground interface, both of which are as wide as the wetted stream width. At least one paper (Dymond 1984) attempts to develop a simple nonenergy balance model that predicts change in temperature based solely on changing the flow and hence width (depth) and time-of-travel. The SNTEMP series of models employs a width as a function of flow relationship in the form of Wea @ where W = width (m)Q = discharge (ems) a and b = empirically derived coefficients It is apparent that this formulation has the following properties. First, if b equals zero, the "a" term becomes the width. Second, the width will be zero if the flow is zero, not accounting for pools. Third, the rel ationship between width and flow is linear if plotted on a log-log scale. The best procedure to develop this relationship is as follows. First, obtain several (three or more) sets of width and flow measurements at random points along each stream segment. This may be accomplished in the field or from output from the HABTAT (or related model such as AVDEPTH) portion of the PHABSIM models, which will report the total stream surface area (per 1,000 feet of stream) as a function of flow on the so-called HAQF output file. Care should be taken to make sure weighting factors are applied to represent the entire segment and that river bends are accounted for if necessary (program ADDBEND, Milhous et al., in press). Second, take the natural log of both width and discharge and perform a standard linear regression with discharge being the independent vartable. The antilog of the intercept should be computed, not forced to zero, because it will be equal to the “a" term in the relationship. The "b" term will be the coefficient (slope) of the regression; the antilog of "b" should not be taken because it is a unitless term. Note that this analysis may be done in any units system you choose as Tong as they are consistent (Figure 5). Appendix C presents the skeleton of a Lotus 1-2-3 worksheet useful for doing this analysis. Some authors (Currier and Hughes 1980) have argued that the width should only be measured for flowing water. Large pools with little or no flow, they state, do not influence the temperature of flowing water. I concur that areas of limited heat interchange may be omitted from width calculations. However, in areas where much of the flow goes through deep pools with little velocity, the width should not be adjusted. What is more important, I believe, is weighting the formulation ‘of the coefficients toward the flow regime of importance. If you know, for instance, that you need the most accurate mode? for low flow conditions, only put low flow width measurements into your regression. Small braided streams will] require more accurate field measurements (Currier and Hughes 1980). If you cannot develop a width-flow relationship, set the b coefficient to zero and employ an average width. ning’ This is a measure of the roughness of the streambed and channel, which causes flowing water to backup due to friction, and is a necessary component of the SNTEMP model in predicting daily maximum water temperatures. At lower flows, the roughness tends to be due primarily to the stream bottom characteristics: as the flow increases, the whole channel shape, including river bends and constrictions, becomes dominant. Therefore, Manning’s n is not constant with 4POUDRE RIVER WIDTH-FLOW RELATIONSHIP English Computation 447 434 42 at 40 LN (Width) (ft) ~ & LN (Discharge) (cfs) POUDRE RIVER WIDTH-FLOW RELATIONSHIP Metric Computation Width = ac? = 13.24Q021 R= 0.98 LN (Width) (m) LN (Discharge) (cms) Figure 5. Development of width versus flow relationships showing no effect of units of measurement on slope, but definite effect on intercept, 15changing flow, even though SNTEMP considers it a constant. Though there are guidelines an experienced hydrologist can use in determining roughness, the confidence interval surrounding such estimates is likely to be large (Platts 1981). Use of a step-backwater hydraulic simulation model, such as the Water Surface Profile (WSP) model (Milhous et al. 1988), may be a better method to estimate n. The use of a regression-type hydraulic model, such as IFG4 (Milhous et al. 1989), however, is not recommended; the "n values" used in this type of model are really “conveyance factors" and not true estimates of channel roughness. Travel Time Travel time is an alternative to Manning’s n. Travel time is the inverse of velocity. If velocity is measured in units of length per time, then travel time is measured in units of time per length, such as seconds per kilometer in the SNTEMP mode]. Stream velocity, and therefore time of travel, vary with discharge. The relationship takes the form: Travel Time = a oe where a and b = empirically derived coefficients Q = discharge Note that the exponent b may itself vary as the flow-control varies with discharge. For example, the stream may change from a fundamental pool-riffle control to a channel control as the discharge increases. Consequently, three or more time-of-travel measurements may be necessary, depending on the range of flows of interest. If no control change takes place, a travel time vs. discharge plot may be constructed (Figure 6). If a control change is evident, such a plot would itself be curvilinear (Hubbard et al. 1981). Travel time may be either estimated or measured for steady or gradually varied flow conditions. Often, travel time estimates are available from power/water companies. If travel time must be estimated from very limited data, the following empirical relationships, adapted from Boning (1974), may be used. These relationships were developed from 873 independent measurements throughout the United States. Note, however, the large standard errors involved. Pool and Riffle Reaches (standard error = 40%) TT=1/ (0.38 g?49) * 0-20 Channel-Controlled Reaches (standard error = 26%) TT = 1/ (2.69 Q°°75) « 59-28 where TT = travel time (s/ft) Q = flow ([Link]./s) S$ = slope of streambed (ft/ft) 16Figure 6. a stream. TRAVELTIME, INHOURS 300 x zor 8 a< 3 150 a 100 =x 90) 2 sor it zor & 60) zB B November 14, 1967 4 June 7, 1968 © September 25, 1968 1 15° 2253 4 5678910 15 20 2530 DISCHARGE, IN CUBIC METERS PER SECOND Variation in travel time versus discharge for selected sites along Reproduced from Hubbard et al. (1981). 17The simplest, but most error prone, method of measuring travel time is the floating object method {Hamilton and Bergersen 1984). Its use is limited to straight and uniform stream segments, with minimum surface waves, on windless days. Floating object is a misnomer, for appropriate objects actually are immersed between one fourth of the depth and the bottom, and do not float on the surface. (In a pinch, a group of oranges may be used.) Establish three to five transects, far enough apart to actually measure an elapsed time. Intermediate transects provide double-checks on the estimates obtained. Several trials will be necessary, with the floats positioned at several locations across the initial transect, with the final answer being the mean time. Finally, multiply the mean velocity of a reasonably smooth stream by 0.8 to obtain the average, midcolumn velocity, If a concurrent or previous PHABSIM study is available for the study stream, the detailed output from one of the hydraulic models may be examined and a mean travel time calculated from the total cross-section area divided by the discharge at each transect. According to one source (Hubbard et al. 1981), this method will tend to underestimate the travel time unless a weighted mean is computed by giving proportional weight to the length of stream represented by each transect (i.e., habitat mapping approach). The next most accurate methods are probably routing studies or colored dye studies. In a routing study conducted below a controlled-release ‘impound- ment, an abrupt increase in flow followed by an abrupt decrease to the previous base flow is made. Staff gages, or stage recorders, located at downstream transects record the sequential passage of the release wave. Travel time is computed from the time of peak stage to peak stage between transects. Different base flows must be used to develop a travel time vs, flow function (Waddie 1987). Colored dye studies involve the instantaneous pouring of fluorescein or potassium permanganate into the stream far enough above the upstream transect to permit complete lateral dispersion, Dye behaves much the same as water molecules and moves on the average at the same rate as water. Travel time is computed by estimating the time when the “center of the color mass" passes the downstream stations. Considerable judgment is usually required to best gage the time at which the "best color" is reached. Experimentation is often necessary to achieve concentrations strong enough to be easily measured, but weak enough to not cause downstream complaints. See Hamilton and Bergersen (1984) for more details. The cadillac of methods is the true fluorometric dye study (Hubbard et al. 1981). The details and equipment are complicated and relatively expensive. A fluorometer is used to measure the light emitted from a fluorescent dye. The dye is selected for properties such as detectability, toxicity, solubility, and cost. The currently recommended dye is rhodamine WT, specifically formulated for water tracing. Concentration-time plots (Figure 7) may be constructed in a detailed dispersion study, or more simple peak-to-peak concentration times may be adequate in a less costly study. Rigorous standards must be met for injecting these dyes into water bodies that have water withdrawal points leading to human consumption. Significant effort is involved in successfully implementing a dye study of this sort. You should seek assistance from a hydrologist experienced in this type of study. 18Site n site n+1 ns Tous Concentration Time f+ tr ie 14 The i Figure 7. Schematic of dye concentration versus time showing dispersion during time-of-travel study. The X-axis also may be interpreted as proceeding downstream from left to right. Thus the magnitude of the concentration becomes attenuated through time (and space). Travel time is measured from ra Soncentration to peak concentration. Reproduced from Hubbard et al. 1981). SNTEMP works with either a constant Manning's n or a constant travel time, both of which are truly dynamic with changes in discharge. If large variations are possible, high and low flow models should be constructed. Thermal Gradient The thermal gradient determines the rate of heat lost or gained from the streambed to the water. The thermal gradient may be thought of as the reciprocal of the more commonly known "r” value used in home insulation. The r value is 19the resistance to heat Toss; the therma} gradient is a measure of the conductance of heat, The larger the difference between the ground temperature and the water temperature, the greater the potential heat transfer. Though determined to be small relative to other parameters (Figures 3 and 4), some authors have determined the thermal gradient to be reasonably sensitive in predicting diurnal temperature variations in small, shallow streams (Jobson 1977). Comparing the results of models run with and without the consideration of a thermal gradient, Jobson determined temperature differences averaging about 0.25 °C (0.45 °F). Comer and Grenney (1977) document a method for assessing the thermal gradient in shallow, sand- and gravel-bed streams. Measurement without disturbance is difficult, but possible. They concluded that for streams with significant interchange of water in the saturated zone below a river, the net heat flux from the "ground" into the stream at night may equal the outgoing flux at the air-water interface. They also suggest that solar radiation may indeed be directly absorbed by the streambed in clear, shallow streams. Very clear streams with black or dark rock bottoms may display different diurnal temperature variation than would be explained by the SNTEMP family of temperature models. Table 3 from Geiger (1965) shows the percentage of incident solar radiation reaching various depths in clear water. Note the significant decline in mid-wavelength radiation between 10 cm and 1m. Thus, bottom conditions may not matter in streams deeper than 10 cm to 1 m. Table 3. Percentage of incident solar radiation reaching various depths in water. Depth Wavelength (u) Imm lem = 10cm Im 10m 100m 0.2-0.6 100.0 100.0 99.7 96.8 72.6 5.9 0.6-0.9 99.8 98.2 84.8 35.8 2.6 0.0 0.9-3.0 65.3 34.7 2.0 0.0 0.0 0.0 In general, my sensitivity analysis demonstrated little influence exerted by the ground temperature. However, Jobson and Keefer (1979) showed that including the heat transfer at the streambed decreased the mean error in their dynamic temperature simulation, but increased the RMS error. They concluded that streambed conduction acts as a “damper” to temperature computation, which, overall, improved their modeling results. Crittenden’s (1978) sensitivity analysis showed that the diffusivity of the streambed was significant in an equilibrium temperature model developed for small, shallow, low gradient streams subject to intense solar radiation. The 20