A Comparison of Statistical Approaches for Predicting Stream Temperatures Across Heterogeneous Landscapes1

Abstract: Estimating stream temperatures across broad spatial extents is important for regional conservation of running waters. Although statistical models can be useful in this endeavor, little information exists to aid in the selection of a particular statistical approach. Our objective was to compare the accuracy of ordinary least‐squares multiple linear regression, generalized additive modeling, ordinary kriging, and linear mixed modeling (LMM) using July mean stream temperatures in Michigan and Wisconsin. Although LMM using low‐rank thin‐plate smoothing splines to measure the spatial autocorrelation in stream temperatures was the most accurate modeling approach; overall, there were only slight differences in prediction accuracy among the evaluated approaches. This suggests that managers and researchers can select a stream temperature modeling approach that meets their level of expertise without sacrificing substantial amounts of prediction accuracy. The most accurate models for Michigan and Wisconsin had root mean square errors of 2.0‐2.3°C, suggesting that only relatively coarse predictions can be produced from landscape‐based statistical models at regional scales. Explaining substantially more variability in stream temperatures likely will require the collection of finer‐scale hydrologic and physiographic data, which may be cost prohibitive for monitoring and assessing stream temperatures at regional scales.     

Regional and Local Scale Modeling of Stream Temperatures and Spatio-Temporal Variation in Thermal Sensitivities

Understanding variation in stream thermal regimes becomes increasingly important as the climate changes and aquatic biota approach their thermal limits. We used data from paired air and water temperature loggers to develop region-scale and stream-specific models of average daily water temperature and to explore thermal sensitivities, the slopes of air–water temperature regressions, of mostly forested streams across Maryland, USA. The region-scale stream temperature model explained nearly 90 % of the variation (root mean square error = 0.957 °C), with the mostly flat coastal plain streams having significantly higher thermal sensitivities than the steeper highlands streams with piedmont streams intermediate. Model R ² for stream-specific models was positively related to a stream’s thermal sensitivity. Both the regional and the stream-specific air–water temperature regression models benefited from including mean daily discharge from regional gaging stations, but the degree of improvement declined as a stream’s thermal sensitivity increased. Although catchment size had no relationship to thermal sensitivity, steeper streams or those with greater amounts of forest in their upstream watershed were less thermally sensitive. The subset of streams with three or more summers of temperature data exhibited a wide range of annual variation in thermal sensitivity at a site, with the variation not attributable to discharge, precipitation patterns, or physical attributes of streams or their watersheds. Our findings are a useful starting point to better understand patterns in stream thermal regimes. However, a more spatially and temporally comprehensive monitoring network should increase understanding of stream temperature variation and its controls as climatic patterns change.     

A Comparison of Protocols and Observer Precision for Measuring Physical Stream Attributes1

Abstract: Stream monitoring programs commonly measure physical attributes to assess the effect of land management on stream habitat. Variability associated with the measurement of these attributes has been linked to a number of factors, but few studies have evaluated variability due to differences in protocols. We compared six protocols, five used by the U.S. Department of Agriculture Forest Service and one by the U.S. Environmental Protection Agency, on six streams in Oregon and Idaho to determine whether differences in protocol affect values for 10 physical stream attributes. Results from Oregon and Idaho were combined for groups participating in both states, with significant differences in attribute means for 9 out of the 10 stream attributes. Significant differences occurred in 5 of 10 in Idaho, and 10 of 10 in Oregon. Coefficients of variation, signal‐to‐noise ratio, and root mean square error were used to evaluate measurement precision. There were differences among protocols for all attributes when states were analyzed separately and as a combined dataset. Measurement differences were influenced by choice of instruments, measurement method, measurement location, attribute definitions, and training approach. Comparison of data gathered by observers using different protocols will be difficult unless a core set of protocols for commonly measured stream attributes can be standardized among monitoring programs.     

Comparison of Machine Learning Models Performance on Simulating Reservoir Outflow: A Case Study of Two Reservoirs in Illinois,U.S.A.

Reservoir outflow is an important variable for understanding hydrological processes and water resource management. Natural streamflow variation, in addition to the streamflow regulation provided by dams and reservoirs, can make streamflow difficult to understand and predict. This makes them a challenge to accurately simulate hydrologic processes at a daily scale. In this study, three Machine Learning (ML) algorithms, Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN), were examined and compared to model reservoir outflow. Past, current, and future hydrologic and meteorological data were used as model inputs, and the outflow of next day was used as prediction. Simulation results demonstrated that all three models can reasonably simulate reservoir outflow. For Carlyle Lake, the coefficient of determination and Nash–Sutcliffe efficiency were each close to one for the three models. The coefficient of determination, relative mean bias, and root mean square error indicated that the SVM performed better than the RF and ANN, but the SVM output displayed a larger relative mean bias than that from RF and ANN. For Lake Shelbyville, the ANN model performed better than RF and SVM when considering the coefficient of determination, Nash–Sutcliffe efficiency, relative mean bias, and root mean square error. The study results demonstrate that the three ML algorithms (RF, SVM, and ANN) are all promising tools for simulating reservoir outflow. Both the accuracy and efficacy of the three ML algorithms are considered to support practitioners in planning reservoir management.     

USING TECHNICAL ADAPTIVE MANAGEMENT TO IMPROVE DESIGN GUIDELINES FOR URBAN INSTREAM STRUCTURES1

ABSTRACT: Adaptive management is a heuristic approach to treating stream restoration projects as continuous, cyclic experiments, yielding results to be incorporated into future decisions. This comprehensive assessment views failures as surprises that are valuable lessons. Monitoring, evaluation of data, and communication of results are critical; the monitoring results trigger feedback mechanisms to invoke adaptation to the newly acquired information and communication of new hypotheses, treatments, or policies. The principles of adaptive management were applied to a monitoring study of three urban stream restoration sites in Maryland. Data were collected and evaluated for various restoration techniques, including vanes, cross vanes, step pools, root wads, imbricated riprap walls, and coir fiber rolls. Improvements to the existing Maryland design guidelines and policies were developed as the feedback mechanism. With the increasing application of adaptive management in stream restoration efforts, it is likely that repeated failures will be prevented and future restoration projects will be more successful in achieving their goals.     

FLOWING WATER RESOURCES1

Using data related to stream order and the morphological characteristics associated with streams of different discharge rates, an estimate of the river resources of the United States is made. The national totals are: 3,200,000 miles total length of rivers; 15,000 square miles of river surface; and 29 cubic miles of water stored in river channels. Using the same techniques, more exact estimates may be made for individual river basins. Suggestions are given for application of the techniques and river data in the management of water resources.     

Simulation of Temperature Mitigation by a Stormwater Detention Pond1

Abstract: A numerical model has been developed to simulate the hydraulic and heat transfer properties of a stormwater detention pond, as part of a simulation tool to evaluate thermal pollution of coldwater streams from stormwater runoff. The model is dynamic (unsteady) and based on principles of fluid mechanics and heat transfer. It is driven by hourly weather data, and specified inflow rates and temperatures. To calibrate and validate the pond model field data were collected on a commercial site in Woodbury, Minnesota. The relationship between pond inflow and outflow rates to precipitation was effectively calibrated using continuously recorded pond levels. Algorithms developed for surface heat transfer in lakes were found to be applicable to the pond with some modification, resulting in agreement of simulated and observed pond surface temperature within 1.0°C root mean square error. The use of an unshaded pond for thermal mitigation of runoff from paved surfaces was evaluated using the pond model combined with simulated runoff from an asphalt parking lot for six years of observed rainfall events. On average, pond outflow temperature was 1.2°C higher than inflow temperature, but with significant event‐to‐event variation. On average, the pond added heat energy to runoff from an asphalt parking lot. Although the pond added total heat energy to runoff, it did reduce the rate of heat outflow from the pond by an order of magnitude due to reductions in volumetric outflow rate compared with the inflow rate. By reducing the rate of heat flow, the magnitude of temperature impacts in a receiving stream were also reduced, but the duration of impacts was increased.     

Field Testing the Riparian Ecosystem Management Model on a Riparian Buffer in the North Carolina Upper Coastal Plain

The riparian ecosystem management model (REMM) was field tested using five years (2005‐2009) of measured hydrologic and water quality data on a riparian buffer located in the Tar‐Pamlico River Basin, North Carolina. The buffer site received NO₃‐N loading from an agricultural field that was fertilized with inorganic fertilizer. Field results showed the buffer reduced groundwater NO₃‐N concentration moving to the stream over a five‐year period. REMM was calibrated hydrologically using daily field‐measured water table depths (WTDs), and with monthly NO₃‐N concentrations in groundwater wells. Results showed simulated WTDs and NO₃‐N concentrations in good agreement with measured values. The mean absolute error and Willmott's index of agreement for WTDs varied from 13‐45 cm and 0.72‐0.92, respectively, while the root mean square error and Willmott's index of agreement for NO₃‐N concentrations ranged from 1.04‐5.92 mg/l and 0.1‐0.86, respectively, over the five‐year period. REMM predicted plant nitrogen (N) uptake and denitrification were within ranges reported in other riparian buffer field studies. The calibrated and validated REMM was used to simulate 33 years of buffer performance at the site. Results showed that on average the buffer reduced NO₃‐N concentrations from 12 mg/l at the field edge to 0.7 mg/l at the stream edge over the simulation period, while the total N and NO₃‐N load reductions from the field edge to the stream were 77 and 82%, respectively.     

COMPARISON OF INSTREAM AND LABORATORY METHODS OF MEASURING SEDIMENT OXYGEN DEMAND1

ABSTRACT: Sediment oxygen demand (SOD) was determined at three sites in a gravel-bottomed central Missouri stream by: (1) two variations of an instream method, and (2) a laboratory method. SOD generally was greatest by the instream methods, which are considered more accurate, and least by the laboratory method. Disturbing stream sediment did not significantly decrease SOD by the instream method. Temperature ranges of up to 12° Celsius had no significant effect on the SOD. In the gravel-bottomed stream, the placement of chambers was critical to obtain reliable measurements. SOD rates were dependent on the method; therefore, care should be taken in comparing SOD data obtained by different methods. There is a need for a carefully researched standardized method for SOD determinations.     

Sensitivity of Thermal Habitat of a Trout Stream to Potential Climate Change,Wisconsin, United States1

Abstract: Airborne thermal remote sensing from four flights on a single day from a single‐engine airplane was used to collect thermal infrared data of a 10.47‐km reach of the upper East Branch Pecatonica River in southwest Wisconsin. The study uses a one‐dimensional stream temperature model calibrated with the longitudinal profiles of stream temperature created from the four thermal imaging flights and validated with three days of continuous stream temperature data from instream data loggers on the days surrounding the thermal remote‐sensing campaign. Model simulations were used to quantify the sensitivity of stream thermal habitat to increases in air and groundwater temperature and changes in base flow. The simulations indicate that stream temperatures may reach critical maximum thresholds for brook trout (Salvelinus fontinalis) and brown trout (Salmo trutta) mortality, particularly if both air temperature increases and base flow declines. The approach demonstrates that thermal infrared data can greatly assist stream temperature model validation due to its high spatial resolution, and that this spatially continuous stream temperature data can be used to pinpoint spatial heterogeneity in groundwater inflow to streams. With this spatially distributed data on thermal heterogeneity and base‐flow accretion, stream temperature models considering various climate change scenarios are able to identify thermal refugia that will be critical for fisheries management under a changing climate.     

LONG-TERM CHANGES IN LOW-FLOW CHANNEL WIDTHS WITHIN THE SOUTH UMPQUA WATERSHED,OREGON1

ABSTRACT: Recent stream survey data (1989–1993) from 31 stream segments of 21 streams within the upper South Umpqua Watershed Oregon were compared to 1937 stream survey data collected from these same stream segments. Current low-flow wetted stream widths of 22 of the 31 surveyed stream segments were significantly different than in 1937; 19 stream segments were significantly wider while the remaining three stream segments were significantly narrower. In only 1 of 8 tributaries to the South Umpqua River which had headwaters within land designated wilderness area did low-flow stream channel width increase since 1937. Conversely, 13 of the 14 tributaries to the South Umpqua River which originated from lands designated as timber emphasis were significantly wider than in 1937. The observed change in stream width was linearly related to timber harvest (r²= 0.44), road density (r²= 0.45), and the amount of large organic debris remaining within the active stream channel (r²= 0.43). These findings suggest that timber harvest and road construction may have resulted in changes in channel characteristics. These channel changes may also be a factor in the observed decline of three of the four populations of anadromous salmonids within the basin.     

Comparative Evaluation of Three River Water Quality Models1

Abstract: Selection of a biochemical oxygen demand (BOD) reaction model to incorporate into dissolved oxygen (DO) water quality models is an overlooked choice available to river water quality modelers. Data from rivers can serve in screening methods to discriminate between competing water quality models. In this study, 15 published BOD and DO datasets based on a 7 year long study of the Bormida River in Italy are used to calibrate three‐multiorder BOD models: first‐order, three‐halves order, and second‐order, which are then included in three corresponding DO models which incorporate these BOD models. The adequacy of the first‐order, three‐halves order and second‐order BOD models was evaluated by calculating the root mean square error between a model and data. A similar procedure was followed to evaluate three DO models, each of which incorporated one of the three BOD models. The first‐order BOD model most frequently fit the river data best, followed by the three‐halves order and the second‐order BOD models. The DO model incorporating a first‐order BOD model most frequently fit the data best, followed by the DO order incorporating second‐order BOD and the DO model incorporating three‐halves order BOD.     

Drying of salted silver jewfish in a hybrid solar drying system and under open sun: Modeling and performance analyses

This study investigated the thin-layer drying kinetics of salted silver jewfish in a hybrid solar drying system and under open sun. Ten drying models were compared with experimental data of salted silver jewfish drying. A new model was introduced, which is an offset linear logarithmic (offset modified Page model). The fit quality of the models was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and sum of squared absolute error (SSAE). The result showed that Midilli et al. model and new model were comparable with two or three-term exponential drying models. This study also analyzed energy and exergy during solar drying of salted silver jewfish. Energy analysis throughout the solar drying process was estimated on the basis of the first law of thermodynamics, whereas exergy analysis during solar drying was determined on the basis of the second law of thermodynamics. At an average solar radiation of 540 W/m² and a mass flow rate of 0.0778 kg/sec, the collector efficiency and drying system efficiency were about 41% and 23%, respectively. Specific energy consumption was 2.92 kWh/kg. Moreover, the exergy efficiency during solar drying process ranged from 17% to 44%, with an average value of 31%. The values of improvement potential varied between 106 and 436 W, with an average of 236 W.     

A METHOD FOR ESTIMATING PEAK WATER DEMAND OF MULTIFAMILY RESIDENCES1

ABSTRACT: The importance of estimating peak water demands for determining the dimensions of pipe size and meters which provide household water to multifamily residences is examined. Several methods utilized in North America and Europe are examined. The analysis makes clear the necessity of studying the peak water demand through statistics based on local data concerning multifamily residences. For different periods of return, the peak demand of a given apartment building is related to its size (the number of apartments) in order to compare the results obtained with existing formula. By use of Ridge-regression technique, the relationship between peak water demand, and building size (number of apartments) and available pressure is established. It can be concluded that peak demand can be estimated with the square root of number of apartment units in the building and the cube root of water pressure.     

FLOODING PROBLEMS IN A SMALL URBAN WATERSHED-DOAN BROOK,CLEVELAND, OHIO1

ABSTRACT: Analysis of a small urban watershed's flooding was undertaken to determine causes and solutions to this serious environmental hazard affecting University Circle, the cultural heart of Greater Cleveland. Doan Brook is a small, highly disturbed urban stream draining 11.3 square miles. Much of the stream coridor and associated park land is owned by the public. The upper watershed lies in the communities of Shaker Heights and Cleveland Heights who lease park land from Cleveland. Two 50-year floods seriously affected the Circle area in August 1975 generating over $1 million in damages. These events resulted from excessive rainfall triggering rapid earth movement of valley walls in the upper watershed, decreased basin lag time from the infilling of several small upland lakes, a seriously undersized stream channel and storm culvert (at University Circle), and complex institutional arrangements between the three communities in the watershed. Suggestions are presented for a methodology to resolve the technical aspects of the flooding problem.     

Predicting Root Density in Streambanks1

Abstract: Roots of riparian vegetation increase streambank erosion resistance and structural stability; therefore, knowledge of root density and distribution in streambanks is useful for stream management and restoration. The objective of this study was to compare streambank root distributions for herbaceous and woody vegetation and to develop empirical models to predict root density. Root length density, root volume ratio, soil physical and chemical properties, and above‐ground vegetation densities were measured at 25 sites on six streams in southwestern Virginia. The Mann‐Whitney test was used to determine differences in root density along stream segments dominated by either woody or herbaceous vegetation. Multiple linear regression was used to develop relationships between root density and site characteristics. Study results showed that roots were evenly distributed across the bank face with the majority of roots having diameters less than 2 mm. Soil bulk density and above‐ground vegetation were key factors influencing root density. While significant relationships were developed to predict root density, the predictive capabilities of the equations was low. Because of the highly variable nature of soil and vegetation properties, it is recommended at this time that soil erodibility and root density be measured in the field for design and modeling purposes, rather than estimated based on empirical relationships.     

Development of a bird integrity index: using bird assemblages as indicators of riparian condition

We describe the development of a bird integrity index (BII) that uses bird assemblage information to assess human impacts on 13 stream reaches in the Willamette Valley, Oregon, USA. We used bird survey data to test 62 candidate metrics representing aspects of bird taxonomic richness, tolerance or intolerance to human disturbance, dietary preferences, foraging techniques, and nesting strategies that were affected positively or negatively by human activities. We evaluated the metric responsiveness by plotting each one against a measure of site disturbance that included aspects of land use/land cover, road density, riparian cover, and stream channel and substrate conditions. In addition, we eliminated imprecise and highly correlated (redundant) metrics, leaving 13 metrics for the final index. Individual metric scores ranged continuously from 0 to 10, and index scores were weighted to range from 0 to 100. Scores were calibrated using historical species information to set expectations for the number of species expected under minimally disturbed conditions. Site scores varied from 82 for the least disturbed stream reach to 8.5 for an urban site. We compared the bird integrity index site scores with the performance of other measures of biotic response developed during this study: a fish index of biointegrity (IBI) and two benthic macroinvertebrate metrics. The three assemblages agreed on the general level of disturbance; however, individual sites scored differently depending on specific indicator response to in-stream or riparian conditions. The bird integrity index appears to be a useful management and monitoring tool for assessing riparian integrity and communicating the results to the public. Used together with aquatic indicator response and watershed data, bird assemblage information contributes to a more complete picture of stream condition.     

Identifying relationships between baseflow geochemistry and land use with synoptic sampling and R-mode factor analysis

The relationship between land use and stream chemistry is often explored through synoptic sampling of rivers at baseflow conditions. However, baseflow chemistry is likely to vary temporally and spatially with land use. The purpose of our study is to examine the usefulness of the synoptic sampling approach for identifying the relationship between complex land use configurations and stream water quality. This study compares biogeochemical data from three synoptic sampling events representing the temporal variability of baseflow chemistry and land use using R-mode factor analysis. Separate R-mode factor analyses of the data from individual sampling events yielded only two consistent factors. Agricultural activity was associated with elevated levels of Ca2+, Mg2+, alkalinity, and frequently K+, SO4(2-), and NO3-. Urban areas were associated with higher concentrations of Na+, K+, and Cl-. Other retained factors were not consistent among sampling events, and some factors were difficult to interpret in the context of biogeochemical sources and processes. When all data were combined, further associations were revealed such as an inverse relationship between the proportion of wetlands and stream nitrate concentrations. We also found that barren lands were associated with elevated sulfate levels. This research suggests that an individual sampling event is unlikely to characterize adequately the complex processes controlling interactions between land use and stream chemistry. Combining data collected over two years during three synoptic sampling events appears to enhance our ability to understand processes linking stream chemistry and land use.     

Large-Scale Effects of Timber Harvesting on Stream Systems in the Ouachita Mountains, Arkansas, USA

Using Basin Area Stream Survey (BASS) data from the United States Forest Service, we evaluated how timber harvesting influenced patterns of variation in physical stream features and regional fish and macroinvertebrate assemblages. Data were collected for three years (1990–1992) from six hydrologically variable streams in the Ouachita Mountains, Arkansas, USA that were paired by management regime within three drainage basins. Specifically, we used multivariate techniques to partition variability in assemblage structure (taxonomic and trophic) that could be explained by timber harvesting, drainage basin differences, year-to-year variability, and their shared variance components. Most of the variation in fish assemblages was explained by drainage basin differences, and both basin and year-of-sampling influenced macroinvertebrate assemblages. All three factors modeled, including interactions between drainage basins and timber harvesting, influenced variability in physical stream features. Interactions between timber harvesting and drainage basins indicated that differences in physical stream features were important in determining the effects of logging within a basin. The lack of a logging effect on the biota contradicts predictions for these small, hydrologically variable streams. We believe this pattern is related to the large scale of this study and the high levels of natural variability in the streams. Alternatively, there may be time-specific effects we were unable to detect with our sampling design and analyses.