Quantifying Representative Sampling Using a Hydrologic Analysis - PowerPoint PPT Presentation

Quantifying Representative Sampling Using a Hydrologic Analysis Tool StormCon 2015 Austin, TX Christian Carleton, PH, CPSWQ, CPESC August 4, 2015 Research Engineer/Hydrologist Office of Water Programs California State University, Sacramento

Outline Introduction Data Review Process Step 1 - Flow Data Quality Step 2 - Sample Collection Timing Step 3 - Percent Capture Conclusion Questions StormCon 2015 Austin, TX August 2-6, 2015

Perspective…and Disclaimer • OWP has over 15 year of stormwater research experience. • Research based monitoring with some regulatory compliance monitoring. • Primarily flow-weighted composite Event Mean Concentration (EMC) water quality sampling. StormCon 2015 Austin, TX August 2-6, 2015

What is representative sampling? Statistics Subset from a population such that the group of samples has the same distribution of characteristics as the entire population. Stormwater Monitoring Flow and water quality data that has the same range and frequency of occurrences as the entire runoff event from a particular location. Locations have the same characteristics as the larger system of interest. StormCon 2015 Austin, TX August 2-6, 2015

Representative Sample Defined in the study plan, Quality Assurance Project Plan (QAPP), Sampling and Analysis Plan (SAP) or similar document. Lots of resources available to plan for and incorporate representative sampling. Not many resources available to determine if a sample is still representative after it has been collected. StormCon 2015 Austin, TX August 2-6, 2015

Data Review Process Step 1 - Flow Data Quality Study plan requirements Collection errors Known Relationships Step 2 - Sample Collection Timing Individual samples taken at appropriate times during event. Step 3 - Percent Capture Quantify how much of the runoff event was sampled. StormCon 2015 Austin, TX August 2-6, 2015

Step 1 – Flow Data Quality Flow-Weighted Sampling Most accurate intra-event sampling scheme. Typically used to obtain Event Mean Concentration (EMC). Error in flow measurements = inappropriate sample timing. Not a true EMC composite sample. Load Calculations Load = EMC x Runoff Volume Error in runoff volume directly translates to error in load calculations. StormCon 2015 Austin, TX August 2-6, 2015

Study Plan Requirements Rainfall Data Runoff Data Duration Duration Depth Time to Peak Volume Peak Flow Rate Intensity Total Volume Average Instantaneous Max 1-hr Max StormCon 2015 Austin, TX August 2-6, 2015

Collection Errors View Raw Data for Physical Constraints Inconsistencies Rainfall Depth v Time • Flow begins after rainfall 0.16 0.14 begins. 0.12 0.1 Depth 0.08 • Flow ends after rainfall 0.06 0.04 ends. 0.02 0 Rainfall Time v Record • Samples collected during 04/05/2010 19:12 04/05/2010 14:24 Time 04/05/2010 09:36 runoff. 04/05/2010 04:48 04/05/2010 00:00 04/04/2010 19:12 Time 04/04/2010 14:24 04/04/2010 09:36 04/04/2010 04:48 04/04/2010 00:00 04/03/2010 19:12 04/03/2010 14:24 165 247 329 411 493 575 657 739 821 903 985 1067 1149 1231 1313 1395 1477 1559 1641 1723 1 83 StormCon 2015 Austin, TX August 2-6, 2015

Known Relationships Volumetric Runoff Coefficient (R v ) 𝑊 𝑠𝑠𝑠𝑠𝑠𝑠 Measured : 𝑆 𝑤 = 𝑊 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (Driscoll 1983) � 𝑤 = 0.858 𝑗 3 − 0.78 𝑗 2 + 0. 774𝑗 + 0.04 Predicted : 𝑆 (Urbonas 1999; WEF and ASCE 1998) Relative Percent Difference (RPD) 𝐽𝐽 𝑆𝑆𝑆 = 𝑆𝑄𝑄𝑄𝑗𝑄𝑄𝑄𝑄 − 𝑁𝑄𝑁𝑁𝑁𝑄𝑄𝑄 ≤ 𝑈𝑈𝑄𝑄𝑁𝑈𝑈𝑈𝑄𝑈𝑁𝑈𝑁𝑄 𝑄 . 𝑕 . , 0.2 𝑄𝑈𝑄𝑢 𝐵𝑄𝑄𝑄𝐵𝑄 𝑆𝑄𝑄𝑄𝑗𝑄𝑄𝑄𝑄 Time of Concentration (T c ) Measured: Time from hyetograph center-of-mass to hydrograph center-of-mass or other method. Predicted: NRCS Method from TR-55 or other method. StormCon 2015 Austin, TX August 2-6, 2015

Step 2 - Sample Collection Timing Check to see if samples were collected at an appropriate time so that they are representative of the runoff. Qualitative - Graphs Quantitative - Uniformity Index StormCon 2015 Austin, TX August 2-6, 2015

Rate Graph Time-Weighted Sampling Runoff Successful Sample Rainfall 0.06 0.0 0.5 0.05 1.0 Rainfall Intensity (in/hr) Runoff Flow Rate (cfs) 0.04 1.5 0.03 2.0 0.02 2.5 0.01 3.0 0.00 3.5 Time StormCon 2015 Austin, TX August 2-6, 2015

Rate Graph Flow-Weighted Sampling Runoff Successful Sample Rainfall 0.06 0.0 0.5 0.05 1.0 Rainfall Intensity (in/hr) Runoff Flow Rate (cfs) 0.04 1.5 0.03 2.0 0.02 2.5 0.01 3.0 0.00 3.5 Time StormCon 2015 Austin, TX August 2-6, 2015

Cumulative Depth Graph Flow-Weighted Sampling 0.30 Rainfall Runoff Successful Sample 0.25 0.20 Depth (in) 0.15 0.10 0.05 0.00 Time StormCon 2015 Austin, TX August 2-6, 2015

Cumulative Depth Graph Time-Weighted Sampling 0.30 Rainfall Runoff Successful Sample 0.25 0.20 Depth (in) 0.15 0.10 0.05 0.00 Time StormCon 2015 Austin, TX August 2-6, 2015

Autosampler Backlog Autosampler sample routine Purge Rinse Sample Purge Routine can take 2+ minutes to complete Urban drainages can have very flashy runoff responses. If the trigger for the next sample is received before the previous sample routine is completed, then it is added to the sample queue. StormCon 2015 Austin, TX August 2-6, 2015

Autosampler Backlog Rate Graph Cumulative Depth Graph 0.35 Runoff Successful Sample Rainfall Rainfall Runoff 0.35 0.0 Successful Sample 0.30 1.0 0.30 2.0 0.25 0.25 3.0 Rainfall Intensity (in/hr) Runoff Flow Rate (cfs) Depth (in) 0.20 4.0 0.20 5.0 0.15 0.15 6.0 0.10 7.0 0.10 8.0 0.05 0.05 9.0 0.00 0.00 10.0 Time Time StormCon 2015 Austin, TX August 2-6, 2015

Qualitative Uniformity Index Want to determine if the intervals are equal. Time or volume interval between samples StormCon 2015 Austin, TX August 2-6, 2015

Uniformity Index Coefficient of Variation (COV) 𝑉𝐽 = 𝐷𝐷𝑈 = 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐸𝐸𝑤𝐸𝑇𝑇𝐸𝐸𝑇 ( 𝜏 ) 𝑁𝐸𝑇𝑇 ( 𝜈 ) COV is a normalized standard deviation Allows for comparison of the variability between different data sets. Small COV -> Little Variation Big COV -> Large Variation Uniformity Threshold (UT) 𝐽𝐽 𝑉𝐽 ≤ 𝑉𝑈 𝑄𝑈𝑄𝑢 𝑉𝑢𝑗𝐽𝑈𝑄𝑉 𝐽𝑢𝑄𝑄𝑄𝐽𝑁𝑈 (0.5 Threshold determined by trial-and error.) StormCon 2015 Austin, TX August 2-6, 2015

Step 3 - Percent Capture • For composite samples. • Many different methods – Percentage of entire runoff time occurring between first and last samples. – Percentage of entire runoff volume occurring between first and last samples. – Etc. StormCon 2015 Austin, TX August 2-6, 2015

OWP Method Assumes flow-weighted sampling 𝑈 𝑇𝐸𝑠 𝑆𝐷 = × 100 𝑈 𝑇𝑠𝑇𝐸𝑠𝑠 V rep = represented volume V runoff = total runoff volume StormCon 2015 Austin, TX August 2-6, 2015

Represented Volume (V rep ) Volume represented in the sample. Best visualized with a cumulative runoff (mass curve) hydrograph. StormCon 2015 Austin, TX August 2-6, 2015

Represented Volume (V rep ) S 6 Volume S 5 S 4 S 3 S 2 S 1 Time StormCon 2015 Austin, TX August 2-6, 2015

Represented Volume (V rep ) ∆ V 6 S 6 ∆ V 5 S 5 Volume ∆ V 4 S 4 ∆ V 3 𝑛 S 3 𝑈 𝑇𝐸𝑠 = � ∆𝑈 ∆ V 2 𝐸 S 2 𝐸=1 ∆ V 1 S 1 Time StormCon 2015 Austin, TX August 2-6, 2015

Represented Volume (V rep ) Δ V 6 S 6 S 6 S 5 Volume S 5 Volume Δ V 4 S 4 S 4 S 3 S 3 Δ V 2 S 2 S 2 Δ V 1 S 1 S 1 Time Time 𝑛 𝑇𝐸𝑠 = ∑ 𝑈 ∆𝑈 = ∆𝑈 1 + ∆𝑈 2 + ∆𝑈 4 + ∆𝑈 6 𝐸 𝐸=1 StormCon 2015 Austin, TX August 2-6, 2015

Total Runoff Volume (V runoff ) Total event runoff volume. Numeric integration method. 𝑇−1 𝑇=𝑇 = � 𝑟 𝑄 + 𝑟 𝑄 + 1 𝑈 𝑇𝑠𝑇𝐸𝑠𝑠 = � 𝑟 𝑄 𝑄 + 1 − 𝑄 2 𝑇=0 𝑇=0 StormCon 2015 Austin, TX August 2-6, 2015

Percent Capture 𝑈 𝑇𝐸𝑠 𝑆𝐷 = × 100 𝑈 𝑇𝑠𝑇𝐸𝑠𝑠 V rep = represented volume V runoff = total runoff volume Minimum Allowable PC If PC ≥ Minimum Allowable them Representative StormCon 2015 Austin, TX August 2-6, 2015

Conclusion Any monitoring project should have a post-data collection QC process. 3-point data review process Flow Data Quality Project Requirements Data Errors Known Relationships Sample Collection Timing Rate Graph Cumulative Depth Graph Uniformity Index Percent Capture Intent is to ensure quality monitoring data is generated, either for research or regulatory compliance purposes. StormCon 2015 Austin, TX August 2-6, 2015

Questions? Thank You Christian Carleton christian.carleton@owp.csus.edu Office of Water Programs http://www.owp.csus.edu