Presentation on theme: "The LOWER DELAWARE MONITORING PROGRAM S WATER QUALITY DATA ANALYSIS PROTOCOL Robert Limbeck Watershed Scientist, DRBC NJ Volunteer Monitoring Summit October."— Presentation transcript:
The LOWER DELAWARE MONITORING PROGRAM S WATER QUALITY DATA ANALYSIS PROTOCOL Robert Limbeck Watershed Scientist, DRBC NJ Volunteer Monitoring Summit October 1, 2004
Lower Delaware River Study Area Control Point Monitoring Concepts
Study Design Objectives Establish baseline Existing Water Quality for future comparison; Set targets to maintain water quality where standards are met; Set targets to improve water quality where standards are not met; Set geographic and water quality priorities to meet the targets; and Monitor long-term to assess trends, prioritize management activities, and assess effectiveness of implementation. How does water quality change from the Delaware Water Gap to Trenton? Which tributaries produce such changes? Where should resources be devoted for most water quality benefit?
QA/QC Considerations Know Your Data: Precision - The degree of agreement of repeated measurements Accuracy - How close your results are to a true or expected value Representativeness - Data represent the true environmental condition Completeness - Compare amount of valid or useable data planned to be collected versus the amount actually collected Comparability - The extent to which data can be compared to other sample locations or periods of time
Data Management Go to www.drbc.net, click on Lower Delawarewww.drbc.net See reports, QAPP, and data files for data management details
The LDMP Excel File Available at www.drbc.net
Always Prepare Metadata Etc.
Data Checking and Cleaning Check accuracy of data entry (multiple checks are best) Check validity of data (can fish live in 1000 degrees F?) Check precision and accuracy of data (QA/QC process) Once in table form, sort all columns, check formats and ranges Make sure blank cells are really blank… Plot data, explain outliers (we rarely exclude outlier data) Quantify non-detect values: we use ½ MDL if <20% non-detects If more than 20% of data are non-detects, they may bias data set The % of non-detects may be a good water quality indicator
Stats Packages Used at DRBC Analyse-ItExcel Add-In from www.analyse-it.comwww.analyse-it.com SASPowerful! www.sas.comwww.sas.com ExcelStatistical functions available, but… There are many other good statistical software applications available. Prices range from $100 to $$$$$.
Representative Data Is your data representative of watershed conditions?
Natural Variability What range of stream flow does your data cover?
Overcome Natural Variability What time of day does your data represent?
Summary Stats – Frequency Plot DO % Saturation at Portland, PA 2000-2003
Summary Stats (Box and Normal Plots) DO % Saturation at Portland, PA 2000-2003 (Data are normally distributed)
Non- Normal Data Enterococcus counts at Portland, PA 2000-2003
Transformed Data Enterococcus counts, Portland, PA 2000-2003, log transformed – produces normality (antilog of results is the geometric mean)
Comparative Stats River mile plot of log Fecal Coliform counts in the Delaware River 2000-2003
Means or Medians? Use means and parametric stats if normality assumptions are met (sample population must be a random sample following the normal distribution). Medians and non-parametric tests make no assumption about the data distribution. Valid for any data set. Note: Parametric tests like the t-test require at least 30 data points for the assumption of normality to be met. Do you collect that much information? In order for parametric comparisons between sites to be made, their data distributions must be the same shape. Do not assume this is the case! Non-parametric stats are safe…
Parametric vs Non-Parametric Tests ParametricNon-Parametric Testing for a Difference: T-TestMann-Whitney U Test 1-Way ANOVAKruskal-Wallis 1-Way ANOVA Testing for a Change: Paired Samples T-TestWilcoxon Signed Ranks Test 1-Way Repeat Measures ANOVAFriedman 1-Way ANOVA Testing for an Association (Correlation): Pearson CorrelationSpearman Rank Correlation Kendall Rank Concordance Prediction: Linear, Deming, or Polynomial RegressionPassing & Bablok Regression
Presentation of Data – Graphs All Data Plot
Presentation of Data – Graphs Medians and Percentiles
Presentation of Data - Tables
Resources USGS Statistical Guidance Manual for Water Resource Studies (USGS manual is excellent…) USEPA guidance (numerous sources, try a web search) Zar, J.H. Biostatistical Analysis. Many others, give me a call… Bob Limbeck, DRBC 609-883-9500 ext 230 Robert.Limbeck@drbc.state.nj.us www.drbc.net