Intervention analysis in environmental engineering

Intervention analysis techniques are described for identifying and statistically modelling trends which may be present in water quality time series. At the exploratory data analysis stage, simple graphical and modelling methods can be employed for visually detecting and examining trends in a time series caused by one or more external interventions. For instance, a plot of a robust locally weighted regression smooth through a graph of the observations over time may reveal trends and other interesting statistical properties contained in the time series. In addition, statistical tests, such as different versions of the nonparametric Mann-Kendall test, can be used to detect the presence of trends caused by unknown or known external interventions. To characterize rigorously and estimate trends which may be known in advance or else detected using exploratory data analysis studies, different parametric methods can be utilized at the confirmatory data analysis stage. Specifically, the time series modelling approach to intervention analysis can be employed to estimate the magnitudes of the changes in the mean level of the series due to the interventions. Particular types of regression models can also be used for estimating trends, especially when there are many missing observations. To demonstrate how intervention analysis methods can be effectively used in environmental impact assessment, representative applications to water quality time series are presented.Invited Paper for Presentation at The Workshop on Statistical Methods for the Assessment of Point Source Pollution, The Canada Centre for Inland Waters, Burlington, Ontario, Canada, L7R 4A6, September 12–14, 1988.     

Modelling Temperature Trends in New Zealand

If global warming is accelerating, then one might expect temperatures for most stations to be accelerating and perhaps variability to be increasing. In this study, we examine 57 New Zealand temperature time series for evidence of non-linearity and changing variability. These correspond to time series for annual minima, annual means and annual maxima for 19 stations. Estimation is by an extended least-squares method. We find a surprising diversity of behaviour of these series – presumably reflecting their different geographic factors as well as series length. We give evidence of regions where temperatures are decreasing. For series where a linear trend is significant, it is downwards in about one third of the cases. This proportion was higher in the South Island, especially for series of minima. Where a non-linear trend is significant, temperatures are decelerating in about one half of the cases. The ratio of downward to upward trends is highest among annual maxima and South Island minima and smallest in annual means. Where a linear trend in the variability is significant, it is decreasing in 13 cases and increasing in 5 cases, although possibly this is partly due to poorer quality data last century. Where a non-linear trend in the variability is significant, variability is decelerating in about two thirds of the cases. The results are used to project upper and lower return levels of minima, means and maxima for each of the series to the year 2010.     

Roadmap for assessing regional trends in groundwater quality

Assessing regional trends in groundwater quality can be a difficult task. Data are often scattered in space and time, and the inertia of groundwater systems can create natural, seemingly persistent changes in concentration that are difficult to separate from anthropogenic trends. Here, we show how statistical methods and software for joint analysis of multiple time series can be integrated into a roadmap for trend analysis and critical examination of data quality. Ordinary and partial Mann–Kendall (MK) tests for monotonic trends and semiparametric smoothers for multiple time series constitute the cornerstones of our procedure. The MK tests include a simple and easily implemented method to correct for serial dependence, and the associated software is designed to enable convenient handling of numerous data series and to accommodate covariates and nondetects. The semiparametric smoothers are intended to facilitate detection of synchronous changes in a network of stations. A study of Swedish groundwater quality data revealed true upward trends in acid-neutralizing capacity and downward trends in sulfate but also a misleading shift in alkalinity level that would have been difficult to detect if the time series had been analyzed separately.     

Long-term variations of water quality parameters in the Maroon River, Iran

Sixteen water quality parameters have been monitored at four stations located along the Maroon River during 1989?C2008. The trend analysis was performed on seasonal and annual time-scales using the Mann?CKendall test, the Sen??s slope estimator and the linear regression. The relationships of the water quality parameters to river discharge were also investigated. The statistical methods showed both positive and negative trends in annual water quality data. However, significant trends were detected by the statistical methods only in calcium, magnesium, sodium absorption ratio, pH, and turbidity series. The results indicated that the concentrations of the water quality parameters increased in spring and winter seasons, while the concentrations were diluted in summer and autumn seasons in the last two decades. Moreover, the highest numbers of significant trends were found in the spring and summer series, respectively. According to the regression analysis, most of the water quality parameters were negatively correlated with river discharge.     

Meteorological normalisation and non-parametric smoothing for quality assessment and trend analysis of tropospheric ozone data

Despite extensive efforts to ensure that sampling and installation and maintenance of instruments are as efficient as possible when monitoring air pollution data, there is still an indisputable need for statistical post processing (quality assessment). We examined data on tropospheric ozone and found that meteorological normalisation can reveal (i) errors that have not been eliminated by established procedures for quality assurance and control of collected data, as well as (ii) inaccuracies that may have a detrimental effect on the results of statistical tests for temporal trends. Moreover, we observed that the quality assessment of collected data could be further strengthened by combining meteorological normalisation with non-parametric smoothing techniques for seasonal adjustment and detection of sudden shifts in level. Closer examination of apparent trends in tropospheric ozone records from EMEP (European Monitoring and Evaluation Programme) sites in Finland showed that, even if potential raw data errors were taken into account, there was strong evidence of upward trends during winter and early spring.     

Hydrological influences on the water quality trends in Tamiraparani Basin,South India

Water quality variables – Turbidity, pH, Electrical Conductivity(EC), Chlorides and Total Hardness (TH) were monitored at adownstream location in the Tamiraparani River during 1978–1992. The observations were made at weekly intervals in a watertreatment and supply plant using standard methods. Graphical andstatistical analyses were used for data exploration, trenddetection and assessment. Box-Whisker plots of annual andseasonal changes in variables indicated apparent trends beingpresent in the data and their response to the seasonal influenceof the monsoon rainfall. Further, the examination of the medianvalues of the variables indicated that changes in the directionof trend occurred during 1985–1986, especially in pH, EC and TH. The statistical analyses were done using non-parametric methods,the ANCOVA on rank transformed data and the Seasonal Man-Kendalltest. The presence of monotonic trend in all the water qualityvariables was confirmed, however, with independent direction ofchange. The trend line was fitted by the method of leastsquares. The estimated values indicated significant increases inEC (28 S cm^-1) while significant decreases were observed inturbidity (90 NTU), pH (0.78), and total hardness (23 ppm) in a span of 15 years. The changes induced in river flow by the addition of a stabilizing reservoir, the influence of seasonal and spatialpattern of monsoon rainfall across the river basin and the increased agriculture appear causative factors for the water quality trends seen in the Tamiraparani River system.     

Stochastic approaches for time series forecasting of boron: a case study of Western Turkey

In the present study, a seasonal and non-seasonal prediction of boron concentrations time series data for the period of 1996–2004 from Büyük Menderes river in western Turkey are addressed by means of linear stochastic models. The methodology presented here is to develop adequate linear stochastic models known as autoregressive integrated moving average (ARIMA) and multiplicative seasonal autoregressive integrated moving average (SARIMA) to predict boron content in the Büyük Menderes catchment. Initially, the Box–Whisker plots and Kendall’s tau test are used to identify the trends during the study period. The measurements locations do not show significant overall trend in boron concentrations, though marginal increasing and decreasing trends are observed for certain periods at some locations. ARIMA modeling approach involves the following three steps: model identification, parameter estimation, and diagnostic checking. In the model identification step, considering the autocorrelation function (ACF) and partial autocorrelation function (PACF) results of boron data series, different ARIMA models are identified. The model gives the minimum Akaike information criterion (AIC) is selected as the best-fit model. The parameter estimation step indicates that the estimated model parameters are significantly different from zero. The diagnostic check step is applied to the residuals of the selected ARIMA models and the results indicate that the residuals are independent, normally distributed, and homoscadastic. For the model validation purposes, the predicted results using the best ARIMA models are compared to the observed data. The predicted data show reasonably good agreement with the actual data. The comparison of the mean and variance of 3-year (2002–2004) observed data vs predicted data from the selected best models show that the boron model from ARIMA modeling approaches could be used in a safe manner since the predicted values from these models preserve the basic statistics of observed data in terms of mean. The ARIMA modeling approach is recommended for predicting boron concentration series of a river.     

Methodological uncertainties in multi-regression analyses of middle-atmospheric data series

Multi-regression analyses have often been used recently to detect trends, in particular in ozone or temperature data sets in the stratosphere. The confidence in detecting trends depends on a number of factors which generate uncertainties. Part of these uncertainties comes from the random variability and these are what is usually considered. They can be statistically estimated from residual deviations between the data and the fitting model. However, interferences between different sources of variability affecting the data set, such as the Quasi-Biennal Oscillation (QBO), volcanic aerosols, solar flux variability and the trend can also be a critical source of errors. This type of error has hitherto not been well quantified. In this work an artificial data series has been generated to carry out such estimates. The sources of errors considered here are: the length of the data series, the dependence on the choice of parameters used in the fitting model and the time evolution of the trend in the data series. Curves provided here, will permit future studies to test the magnitude of the methodological bias expected for a given case, as shown in several real examples. It is found that, if the data series is shorter than a decade, the uncertainties are very large, whatever factors are chosen to identify the source of the variability. However the errors can be limited when dealing with natural variability, if a sufficient number of periods (for periodic forcings) are covered by the analysed dataset. However when analysing the trend, the response to volcanic eruption induces a bias, whatever the length of the data series. The signal to noise ratio is a key factor: doubling the noise increases the period for which data is required in order to obtain an error smaller than 10%, from 1 to 3-4 decades. Moreover, if non-linear trends are superimposed on the data, and if the length of the series is longer than five years, a non-linear function has to be used to estimate trends. When applied to real data series, and when a breakpoint in the series occurs, the study reveals that data extending over 5 years are needed to detect a significant change in the slope of the ozone trends at mid-latitudes.     

Statistical models for evaluating suspected artefacts in long-term environmental monitoring data

Long-term water quality monitoring is of high value for environmental management as well as for research. Artificial level shifts in time series due to method improvements, flaws in laboratory practices or changes in laboratory are a common limitation for analysis, which, however, are often ignored. Statistical estimation of such artefacts is complicated by the simultaneous existence of trends, seasonal variation and effects of other influencing factors, such as weather conditions. Here, we investigate the performance of generalised additive mixed models (GAMM) to simultaneously identify one or more artefacts associated with artificial level shifts, longitudinal effects related to temporal trends and seasonal variation, as well as to model the serial correlation structure of the data. In the same model, it is possible to estimate separate residual variances for different periods so as to identify if artefacts not only influence the mean level but also the dispersion of a series. Even with an appropriate statistical methodology, it is difficult to quantify artificial level shifts and make appropriate adjustments to the time series. The underlying temporal structure of the series is especially important. As long as there is no prominent underlying trend in the series, the shift estimates are rather stable and show less variation. If an artificial shift occurs during a slower downward or upward tendency, it is difficult to separate these two effects and shift estimates can be both biased and have large variation. In the case of a change in method or laboratory, we show that conducting the analyses with both methods in parallel strongly improves estimates of artefact effects on the time series, even if certain problems remain. Due to the difficulties of estimating artificial level shifts, posterior adjustment is problematic and can lead to time series that no longer can be used for trend analysis or other analysis based on the longitudinal structure of the series. Before carrying out a change in analytic method or laboratory, it should be considered if this is absolutely necessary. If changes cannot be avoided, the analysis of the two methods considered, or the two laboratories contracted, should be run in parallel for a considerable period of time so as to enable a good assessment of changes introduced to the data series.     

Trends in Nitrogen Transport in Swedish Rivers

Concern about nitrogen loads in marine environments has drawn attention to the existence and possible causes of long-term trends in nitrogen transport in rivers. The present study was based on data from the Swedish environmental monitoring programme for surface water quality; the continuity of these data is internationally unique. A recently developed semiparametric method was employed to study the development of relationships between runoff and river transport of nitrogen since 1971; the observed relationships were then used to produce time series of flow-normalised transports for 66 sites in 39 river basins. Subsequent statistical analyses of flow-normalised data revealed only few significant downward trends (p 0.05) during the time period 1971–1994, and the most pronounced of these downward trends were caused by reduced point emissions of nitrogen. The number of significant upward trends was substantially larger (15 for total-N and 18 for NO₃-N). Closer examination of obtained results revealed the following: (i) the most pronounced upward trends were present downstream of lakes, and (ii) observed increases in nitrogen transport coincided in time and space with reduced point emissions of phosphorus or organic matter. This indicated that changes in the retention of nitrogen in lakes were responsible for the upward nitrogen trends. The hypothesis that nitrogen saturation of forest soils has caused a general increase in the riverine export of nitrogen from forested catchments in Sweden was not confirmed. Neither did the results indicate that improved agricultural practices have reduced the export of nitrogen from agricultural catchments.     

Monitoring of the Endangered Pinna nobilis Linné, 1758 in the Mar Grande of Taranto (Ionian Sea, Italy)

Two methods were used to calculate the meteorologically adjusted ground level ozone trends in southern Taiwan. The first method utilized is a robust linear regression method. The second approach uses a multilayer perceptron (MLP) artificial neural network (ANN) method. The observations obtained from 16 monitoring stations were analyzed and divided into six groups by hierarchical divisive clustering procedure. The daily maximum 1 and 8 h ozone concentrations for each group are then calculated. The meteorologically adjusted trends obtained by linear regression and MLP methods are smaller than the unadjusted trends for all groups and average time. It indicts that the meteorological conditions in Taiwan tend to increase ambient ozone concentrations in recent years.     

A Method for Ensemble Wildland Fire Simulation

An ensemble simulation system that accounts for uncertainty in long-range weather conditions and two-dimensional wildland fire spread is described. Fuel moisture is expressed based on the energy release component, a US fire danger rating index, and its variation throughout the fire season is modeled using time series analysis of historical weather data. This analysis is used to characterize the seasonal trend in ERC, autocorrelation of residuals, and daily standard deviation and stochastically generate artificial time series of afternoon fuel moisture. Daily wind speed and direction are sampled stochastically from joint probabilities of historical wind speed and direction for the date range of the fire simulation period. Hundreds or thousands of fire growth simulations are then performed using the synthetic fire weather sequences. The performance of these methods is evaluated in terms of the number of ensemble member simulations, one- versus two-dimensional fire spread simulations, and comparison with results from 91 fires occurring from 2007 to 2009. Simulations were found to be in consistent agreement with observations, but trends indicate that the ensemble average of simulated fire sizes were consistently larger than actual fires whereas the farthest extent burned by fires was underestimated.     

Towards generalised reference condition models for environmental assessment: a case study on rivers in Atlantic Canada

Evaluation of the ecological status of river sites in Canada is supported by building models using the reference condition approach. However, geography, data scarcity and inter-operability constraints have frustrated attempts to monitor national-scale status and trends. This issue is particularly true in Atlantic Canada, where no ecological assessment system is currently available. Here, we present a reference condition model based on the River Invertebrate Prediction and Classification System approach with regional-scale applicability. To achieve this, we used biological monitoring data collected from wadeable streams across Atlantic Canada together with freely available, nationally consistent geographic information system (GIS) environmental data layers. For the first time, we demonstrated that it is possible to use data generated from different studies, even when collected using different sampling methods, to generate a robust predictive model. This model was successfully generated and tested using GIS-based rather than local habitat variables and showed improved performance when compared to a null model. In addition, ecological quality ratio data derived from the model responded to observed stressors in a test dataset. Implications for future large-scale implementation of river biomonitoring using a standardised approach with global application are presented.     

Quantitative Ethnobotany in an Atlantic Forest Fragment of Northeastern Brazil – Implications to Conservation

Two new methods for assessing temporal trends in stream-solute concentrations at specific streamflow ranges were applied to long (40 to 50-year) but sparse (bi-weekly to quarterly sampling) stream-water quality data collected at three forested mesoscale basins along an atmospheric deposition gradient in the northeastern United States (one in north-central Pennsylvania, one in southeastern New York, and one in eastern Maine). The three data sets span the period since the implementation of the Clean Air Act in 1970 and its subsequent amendments.Declining sulfate (O ₄ ^2-) trends since the mid 1960s were identified for all 3 rivers by one or more of the 4 methods of trend detection used. Flow-specific trends were assessed by segmenting the data sets into 3-year and 6-year blocks, then determining concentration-discharge relationships for each block. Declining sulfate (O ₄ ^2-) trends at median flow were similar to trends determined using a Seasonal Kendall Tau test and Sen slope estimator. The trend of declining O ₄ ^2- concentrations differed at high, median and low flow since the mid 1980s at YWC and NR, and at high and low flow at WR, but the trends leveled or reversed at high flow from 1999 through 2002. Trends for the period of record at high flows were similar to medium- and low-flow trends for Ca²⁺+ Mg²⁺ concentrations at WR, non-significant at YWC, and were more negative at low flow than at high flow at NR; trends in nitrate (NO₃ ^-), and alkalinity (ALK) concentrations were different at different flow conditions, and in ways that are consistent with the hydrology and deposition history at each watershed. Quarterly sampling is adequate for assessing average-flow trends in the chemical parameters assessed over long time periods (∼decades). However, with even a modest effort at sampling a range of flow conditions within each year, trends at specified flows for constituents with strong concentration-discharge relationships can be evaluated and may allow early detection of ecosystem response to climate change and pollution management strategies.     

用稳健统计法处理方法标准验证实验中方法检出限数据

方法检出限是分析方法基本性能参数之一。根据《环境监测分析方法标准制修订技术导则》(HJ 168—2010)的要求,此参数需要6家以上实验室基于特定样品测试数据的标准偏差得到。目前,标准文本大多采用各家实验室方法检出限中最大值。研究提出用稳健统计法处理方法验证数据,无需识别和删去离群值,可将离群值的影响降到最小。基于4组公开发表数据和2组实测数据,比较了不同方法计算结果。结果表明:参与统计的3组数据29个检测项目中,常规方法确认的方法检出限被判为离群值占27.6%。取最大值存在数值偏高的风险。提出了用曼德尔k检验法识别1组方法检出限数据中的离群值,该方法同样可以用于实验室数据与标准文本中方法检出限符合性判断。     

Monitoring fluvial water chemistry for trend detection: hydrological variability masks trends in datasets covering fewer than 12 years

This paper sub-samples four 35 year water quality time series to consider the potential influence of short-term hydrological variability on process inference derived from short-term monitoring data. The data comprise two time series for nitrate (NO(3)-N) and two for DOC (using water colour as a surrogate). The four catchments were selected not only because of their long records, but also because the four catchments are very different: upland and lowland, agricultural and non-agricultural. Multiple linear regression is used to identify the trend and effects of rainfall and hydrological 'memory effects' over the full 35 years, and then a moving-window technique is used to subsample the series, using window widths of between 6 and 20 years. The results suggest that analyses of periods between six and eleven years are more influenced by local hydrological variability and therefore provide misleading results about long-term trends, whereas periods of longer than twelve years tend to be more representative of underlying system behaviour. This is significant: if such methods for analysing monitoring data were used to validate changes in catchment management, a monitoring period of less than 12 years might be insufficient to demonstrate change in the underlying system.     

Robust Regression with Data-Dependent Regularization Parameters and Autoregressive Temporal Correlations

We introduce robust procedures for analyzing water quality data collected over time. One challenging task in analyzing such data is how to achieve robustness in presence of outliers while maintaining high estimation efficiency so that we can draw valid conclusions and provide useful advices in water management. The robust approach requires specification of a loss function such as the Huber, Tukey’s bisquare and the exponential loss function, and an associated tuning parameter determining the extent of robustness needed. High robustness is at the cost of efficiency loss in parameter loss. To this end, we propose a data-driven method which leads to more efficient parameter estimation. This data-dependent approach allows us to choose a regularization (tuning) parameter that depends on the proportion of “outliers” in the data so that estimation efficiency is maximized. We illustrate the proposed methods using a study on ammonium nitrogen concentrations from two sites in the Huaihe River in China, where the interest is in quantifying the trend in the most recent years while accounting for possible temporal correlations and “irregular” observations in earlier years.     

Qualitative time trend analysis of ground water monitoring networks

The detection of significant (short-term) time trends is one of the major goals of ground water monitoring networks. These trends can be used to recognize active geochemical processes and potential environmental threats. This paper presents a case history of time trend analysis on macrochemical parameters of ground water quality. It shows the difficulties and traps that are generally encountered in such studies. The data used originated from the Dutch National Groundwater Quality Monitoring Network. This network is operative since 1979, and keeps track of the ground water composition at 350 locations at two depths (ca. 10 and 25 m below surface; general density, one location per 100 km²). Prior to the trend analysis the data set was divided into geochemically homogeneous groups using fuzzy c-means clustering. Each group represents a specific ground water type, characterized by a distinct source (seawater, surface water or precipitation) and a unique combination of dominant geochemical processes (e.g. mineralization of organic matter, carbonate dissolution and cation exchange).To study trends qualitatively, the concentrations of the various macro-constituents in ground water are correlated with time of sampling. The nonparametric and outlier insensitive Spearman rank correlation coefficient is computed per well screen. A frequency distribution of correlation coefficients is formed by combining the Spearman correlation coefficients of all individual wells within a homogeneous group. This distribution is tested for trends against the appropriate theoretical distribution of zero correlation by use of the Kolmogorov-Smirnov one-sample test. The type of trend is derived from the shape of the distribution.Most ground water types show statistically significant qualitative trends, of which many, however, are caused by changes in the sampling and analytical procedures over the monitoring period. After elimination of differences in limits of detection for NO₃, total-P, and NH₄, most trends in these compounds disappeared. In some water types trends for alkalinity, apparent trends for pH, Ec, and total-P are caused by variations in the laboratory practice, e.g. varying storage procedures, leading to erroneous analyses. Other parameters showed statistically significant trends, related to geochemical processes.The most interesting and most substantial trends are observed in the water type characterized by infiltrating rainwater with agricultural pollutants. In this water type the lowering ground water table induces lower rates of evapotranspiration, giving lower concentrations in time of conservative parameters (Cl, Na, Ca). The aerated zone is enlarged, resulting in increased oxidation of organic material, less efficient nutrient (NO₃, K) uptake by plant roots, leading to increased ground water concentrations of nutrients. In other water types trends are quantitatively small. However, trends are not necessarily linear, and all should be closely monitored in future.     

Temporal and Spatial Variation of Nutrient Levels in the Nemunas River (Lithuania and Belarus)

There are discordant results on trends in nutrient river water quality from the economical transition countries in Europe. The present study assessed the impact of these economical changes on the load and concentration at 17 monitoring stations along the Nemunas River and its major tributaries (Lithuania and Belarus). Three time periods were evaluated: the Soviet rule command system period 1986–1991, the transfer to market economy period 1992–1996 and the post reform period 1997–2002. The most surprising result in this study, was the increased area-specific load of NO₃-N from the first to the third period at almost all the sampling sites. The increase was particularly large (43–78%) at the sites in the Lithuanian part of the river. The corresponding load increase in the Belarussian part of the river was only 1–15%. The statistical analyses of concentration data confirm the strong upward NO₃-N trend at the Nemunas mouth and at 5 of the 6 tributaries in the lower part of Nemunas. Temporal and spatial analysis of nitrates transport in the Nemunas River and its main tributaries revealed that nitrates mainly originate from agricultural areas. The upward trends were most likely an effect of ploughing of pastures and unbalanced crop fertilisation in combination with large storage and accumulation of soil-nitrogen during the Soviet period.On contrary to nitrate-N, the area-specific load of PO₄-P decreased significantly from the first to the third period at all sites along the Nemunas River (31–86%). Seasonal (SMK) and Partial (PMK) Mann-Kendall tests on PO₄-P concentrations also showed significant downward trend at 14 of 16 investigated sites. The decrease of PO₄-P levels was attributed to the reduction of municipal and industrial point source emissions and to the decreased livestock numbers.The NH₄-N load showed the same pattern as PO₄-P. At the river mouth the load was 90 kg km⁻² yr⁻¹ during the first period compared to only 20–30 kg km⁻² yr⁻¹ in the third period. The trend test on NH₄-N concentrations detected significant downward trends at 5 out of 16 sites. The declines were explained by decreased emissions from cities and large animal breeding farms.This study showed that trend analysis at multiple sites in a river basin is crucial for the understanding of the variability in time and space. Such analysis is also important for our interpretation of underlying sources and fluxes in a drainage basin over time. This is particularly important for compounds that have different source origin.