1. CCAMP Approach to Healthy Watersheds Report Cards: a Look under the Hood Karen R. Worcester, California Central Coast Water Board David M. Paradies, Central Coast Ambient Monitoring Program Dr. John W. Hunt, U.C. Davis and CCAMP

2. Agenda for Today  Project Overview  Scoring data – CCME Water Quality Index – Our modifications Sample size considerations Oddball parameters Comparability with CCME – Multiple thresholds – Aggregating into sub-indices and indices  Thresholds (or the lack thereof…)  Expanding the toolkit – Change – New data types – Different underlying maps – New statistical approaches  CCAMP data flow, geospatial framework, and open source environment  Support and transferability

3. Project Overview

4. What do water quality managers and decision- makers need from their data?  Where is the problem?  What is causing the problem?  What land uses are associated with the problem?  Where are our best places, that need to be protected?  Where are places that could be enhanced or improved?  Are things getting better or worse? Where??

5. We can answer these types of questions in an assessment report But can we answer them with an online tool that updates as the data does??

6. We can…!

7. 7 Our Vision for the Central Coast… Healthy Watersheds By 2025: Healthy Aquatic Habitat - 80% of aquatic habitat is healthy; remaining 20% exhibit positive trends in key parameters Proper Land Management - 80% of land is managed to maintain proper watershed functions; remaining 20% exhibit positive trends in key parameters Clean Groundwater - 80 percent of ground water is clean, and the remaining 20 percent will exhibit positive trends in key parameters

8. To assess our goals we needed to characterize both status (health) and change  Multi-metric approach  Measured and modeled data  Consistent, threshold-based scoring approach  Status and change at different scales – Analyte and multi-metric scales – Site, waterbody, and watershed scales

9. General principles  Help the user answer Where, Why, What?  Data from readily available online sources  Data of documented quality  Transparency of methods  Drill down for detail  Staff-maintained technical content via wikis

10. Healthy Watersheds Web Report Card, publically available later this year

16. Report Card will connect to CCAMP Data Navigator to access data, maps, graphs, summary stats, trend analysis and other statistical tools

17. Change analysis at San Simeon Creek

19.  Adapted from Canadian Water Quality Index (CCME)  Magnitude and exceedance components  “Report card” paradigm We are also designating Outstanding (A+) for “Blue Water Streams” that score Excellent across all measures. Scoring Approach 5 Excellent Good Fair Poor Very Poor

20. Thresholds  We have compiled many water and sediment thresholds into a single table  Software assisted threshold selection  Threshold table can be adapted for other purposes.  Scaling approach for multiple thresholds

21. Scoring whole watersheds (currently in progress)

22. Flow and Loading Assign scores to upstream reaches using NHDPlus stream network. Use modeled data to score unmonitored areas

23. Modeled data from California’s recent Healthy Watersheds (CADMUS) Assessment

24. Cadmus Stream Health  Cadmus Stream Health layer serves as baseline condition for Region (aquatic life only)  Scoring is regionalized (highest score in Region is set at 100)  Scores are redistributed to 6 categories  Where measured data is available, measured scores are combined Cadmus scores

25. Next Steps Public release of Data Navigator in September. Methods manual to the SWAMP program for peer review this fall. Public release of Phase 1 of the Central Coast Healthy Watersheds Report Card this winter

26. Phase 2 of the Healthy Watersheds Report Card Add linked groundwater data from GeoTracker to Human Health Goal Address Goal 2 related to watershed management Pesticide applications Impervious surfaces Ag program metrics Stormwater program metrics etc.

27. Digging into the details…

28. Scoring Data

29.  Adapted from Canadian Water Quality Index (CCME)  Magnitude and exceedance components  “Report card” paradigm We are also designating Outstanding (A+) for “Blue Water Streams” that score Excellent across all measures. Scoring Approach 5 Excellent Good Fair Poor Very Poor

30. Canadian CCME Water Quality Index (WQI) has three factors Factor 1: Scope Factor 2: Frequency (Exceedance) Factor 3: Amplitude (Magnitude)

33. Calculating CCME WQI

34. MEQ: modifications to CCME  Score all tests for magnitude, not just failed tests. This provides more differentiation at the “good” end of the scoring tool  Eliminate the scope term (percent of variables that fail) and use a different approach for combining parameters  Special handling of some variables

35. Calculating MEQ

36. Factor 3: Magnitude (Amplitude ) Replacing “Failed Test Value” with “Test Value” better differentiates “good” from “excellent”

37. Calculating MEQ, cont. Combine exceedance and magnitude factors and scale to 100

38. Eliminating the Scope Term  Scores at the analyte level instead of the site level, which supports our web report card approach  Lengthy analyte lists with many “non-detects” do not average out the signal (we didn’t want to limit the use of data to a fixed “short list”)  Combining approach can be tailored to analyte type (mean average, harmonic mean, worst score)

39. Comparing MEQ scores to an independent multi-threshold scoring approach We used an entirely different scoring system to verify that the CCME grade breakpoints still made sense using our modified “MEQ” approach

40. Multiple threshold scoring approach used by the original CCAMP website CCAMP “Rule” for Nitrate-N to protect for aquatic life uses: If the 90 th percentile > 2 then The_Color = Dark Red If the 90 th percentile <= 2 Then The_Color = Red If the 90 th percentile <= 1 Then The_Color = Yellow If the 90 th percentile <= 0.3 Then The_Color = Green If the 90 th percentile <= 0.15 Then The_Color = Dark Green The MEQ threshold is set at 1.0 mg/L

41. CCAMP “Rule” for Nitrate-N to protect for domestic use: If the 90 th percentile > 20 then The_Color = Dark Red If the 90 th percentile <= 20 Then The_Color = Red If the 90 th percentile <= 10 Then The_Color = Yellow If the 90 th percentile <= 5 Then The_Color = Green If the 90 th percentile <= 2.5 Then The_Color = Dark Green The MEQ threshold is set at 10.0 mg/L

42. Excellent performance between scoring approaches for some variables (y-axis is number of sites)

43. Poorer performance for others: CCAMP “Rule”: 90 th percentile >= 40 90 th percentile < 40 90 th percentile < 15 90 th percentile < 4.85 90 th percentile < 1.78 MEQ threshold is 15.0 ug/L CCAMP “Rule”: 90 th percentile >= 252 90 th percentile < 252 90 th percentile < 126 90 th percentile < 63 90 th percentile < 32 MEQ threshold is 126 MPN/100 ml In some cases, the CCAMP Rule Performance may be at issue

44. Comparison of Biostimulatory Index using two scoring approaches (colors are based on MEQ score) Index score based on CCAMP Rules Index score based on MEQ

45. Sample Size Considerations

46. When n= 1, there are no intermediate scores because the exceedance term acts like an “on-off” switch

48. Oddball Parameters

49. Exponential distributions Parameters such as coliform, turbidity, TSS In the Magnitude term, we are using a geomean instead of an arithmetic mean; this improves the fit relative to the CCAMP Rules and prevents excessive influence of high values: CCAMP Rule for fecal coliform Geomean >= 400 Geomean < 400 Geomean < 200 Geomean < 100 Geomean < 50

50. Double-ended thresholds pH and Dissolved oxygen (for our Region) Calculate MEQ for each threshold as though for separate parameters; overall score is the lower of the two scores. pH <7Dissolved oxygen <7.0 mg/L pH >8.5Dissolved oxygen >13.0 mg/L

51. Multi-threshold scoring systems So. Cal. IBI CCME Breakpoints

52. Comparability with CCME  The CCME was derived to support compliance, as opposed to a full spectrum health evaluation – It relies on a fixed analyte list – It disregards magnitude of scores under the threshold so has low resolution for characterizing better quality sites – It doesn’t provide for variable combining approaches based on analyte group characteristics

53. MEQ scoring compared to CCME  Placeholder slide

54. BiologyHabitat INDICES OF HEALTH

55. Indices of Health Human Health Index Drinking water Nitrogen species Salts Metals Organic Chemicals Water Contact Pathogens Aquatic Life Index Conventional Analytes Toxicity Biology Biostimulatory Risk Metals Organic Chemicals Habitat

56. Aquatic Life Index Conventional water quality  pH departure  Water temperature  Nitrate - N  Total and unionized ammonia  Orthophosphate - P  Total suspended solids  Turbidity Pesticides and other Organics sediment and water Metals sediment and water Biostimulation  Oxygen departure  Chlorophyll a (ug/L)  % floating mats  NNE oxygen deficit  NNE predicted benthic chlorophyll biomass Toxicity Algal cell growth Fish survival Fish growth Invert survival in water Invert reproduction in water Invert survival in sediment

57. Habitat (future)  Regionally-scaled riparian assessment using imagery analysis in combination with field measures (Central Coast Wetlands Group)  Riparian habitat measures (CRAM and/or pHab)  Instream habitat measures (CRAM and/or pHab) Biology  Benthic invertebrates  Soft-bodied algae  Periphyton  Fish  Amphibians, etc. Aquatic Life Index, cont.

58. Human Health Index DRINKING WATER Nitrogen Species Nitrate Ammonia Nitrite Salts Boron Chloride Sodium TDS Pesticides and other Organics sediment and water Metals sediment and water WATER CONTACT Pathogens E. coli Fecal coliform

59. Aggregating scores into an index Overall Aquatic Life - Harmonic Mean Conventional analytes – Arithmetic Mean Biostimulation – Arithmetic Mean Metals – Harmonic Mean Toxicity – Worst score Organic chemicals – Worst score Overall Human Health – Harmonic mean or worst score of toxic chemicals, whichever is lower Pathogens, salts – Harmonic mean Nitrogen species, Organic chemicals, metals – Worst score

60. Thresholds

61. The complexities of thresholds  Appropriate thresholds are key to the several projects discussed today  Different thresholds apply in different contexts  Scale is important – State – Region – Waterbody – Site

62. Threshold Selection  Assembled thresholds from various sources – Key sources are U.S. EPA Aquatic Life Benchmarks and California Water Quality Goals database  Currently focused on sediment and water only, for human health and aquatic life  Established selection criteria – Marshack algorithms – Health not harm (“threshold” effects) – Consistency within chemical group

63. We adapted the Marshack “algorithms” to a automated approach to threshold selection (see handout)

64. Goal Two - Watershed Management

65. Cadmus Stream Health  Cadmus Stream Health layer serves as baseline condition for Region (aquatic life only)  Scoring is regionalized (highest score in Region is set at 100)  Scores are redistributed to 6 categories  Where measured data is available, measured scores are combined Cadmus scores

66. Watershed stressors and management activities Example management activities: Low impact development Stormwater BMPs Agricultural BMPs Area of drip irrigation Area of organic/no till/ sustainable agriculture Restoration projects

67. Using Cadmus Healthy Watersheds data Stream Health California Stream Condition Index Habitat Index (PHAB and CRAM) Water Quality Index (conductivity, nitrate and turbidity medians) Watershed Condition Percent Natural Cover Percent Intact Active River Area Sedimentation Risk Percent Artificial Drainage Area Dam Storage Ratio Road Crossing Density

68. Expanding the Tool Kit

69. Scoring Change “80% of aquatic habitat is healthy; remaining 20% exhibit positive trends in key parameters”

70. From our website: Nitrate in the Monterey Area

72. (note arrow icons denoting change).

73. This site is crossing a Grade (color) boundary from “Fair” to “Good” (threshold is 1000 mg/L) How did we go from overall scoring to change scoring??

74. We look at change at the site level in two ways: Kendall Trend Analysis

75. We look at change in two ways: Kendall Trend Analysis

76. Some change doesn’t fit a straight line:

79. Change Point Analysis defines probable change points in a time series of data In this case, a treatment plant upgrade went online in May, 2007

80. Apply MEQ scoring to data on each side of Change Point to grade (color) two sections of arrow icon We have found Change Point Analysis to be more useful than traditional trend analysis and are relying on it as our primary change scoring approach.

81. Analyte 4 Analyte 5 Analyte 6 41 7885 67 59 46 41 22 73 88 EXAMPLE: Down arrow = Getting Worse Up Arrow = Getting Better Analyte 3 Analyte 2 Analyte 1 Of six analytes that make up an index, 3 are getting worse, 1 is getting better and two show no change MEQ Grading Key

82. Analyte 4 Analyte 5 Analyte 6 41 7885 67 59 46 41 22 73 88 Using the most recent significant change point for each analyte and the appropriate aggregation approach: Before period: Mean (78 + 59 + 85 + 22 + 73 + 88) = 68 After period: Mean (41 + 46 + 67 + 41 + 73 + 88) = 59 At the level of the index, the site is getting worse Down arrow = Getting Worse Up Arrow = Getting Better 59 68 Analyte 3 Analyte 2 Analyte 1

83. Software framework

84. What is CCAMP OpenWater?

85. CCAMP OpenWater is… …An Internet-based Open Source Software Toolkit focused on water quality and quantity assessment and visualization. "If I can't picture it, I can't understand it." (Albert Einstein) Multi-server environment Scheduled data mining from multiple databases Data grooming Statistical Analysis Data visualization tools

86. Why Open Source?  Reduces system development failure risk  Provides access to international community of code developers and standards  Empowers agency staff, users, and development partners  Avoids pre-committing users to licensing agreements with sole source commercial vendors

87. Current Data Flow CEDEN (future) GeoTracker CDPR Pesticide Use Report SWAMP USGS NWIS SWRCB Water Quality Goals CCAMP Staging Data Tables Web Views The system is intended to use routine automated queries to keep data up-to-date.

88. Data Grooming Synonym dictionaries Analyte name standardization Units of measurement standardization Quality assurance data filtering Handling of duplicates

89. GeoSpatial processing and linking Automated linking of monitoring sites to GIS layers Handling GIS layer idiosyncrasies Linking of land use and other datasets to sites Pesticide use characterization Land Cover characterization Flow and Load estimation

90. Geospatial Framework  National Hydrography Dataset Plus  National Watershed Boundary Dataset  National Land Cover Dataset  Public Land Survey System Boundaries  Bulletin 118 Groundwater Basin Boundaries  California Healthy Watersheds (CADMUS)

91. Current Data Sources  CCAMP, SWAMP, Central Coast Ag data, MPSL Grant data  CEDEN (aka SWAMP Data Warehouse Expanded)  GAMA/Geotracker  CDPR pesticide use database  Cadmus data layers  National Land Cover Dataset  USGS Flow gage data  National Watershed Boundary Dataset  National Hydrography Dataset Water quality data types include field data, flow, water and sediment chemistry, pathogen indicators, water and sediment toxicity, bioassessment, NNE model outputs

92. New Data Types  If threshold can be assigned apply MEQ scoring approach and check performance  If a multiple threshold approach exists, redistribute scores as appropriate to create A, B, C, D, and F categories  If no threshold exists, consider a quantile approach.

93. Different underlying maps A simple shapefile format is all that is needed:  'ESRI shape files' are default gis data exchange format.  Shape Files with viable shapefilename.prj element can be added  We may in the future include 'web services' based layers.

94. Other statistical approaches We maintain an open source Library of statistical functions and packages.  Database driven batch production of: – descriptive statistics – graphics to examine statistical assumptions and to provide exploratory data analysis – Bivariate and multivariate analysis  Future: Calculation of CSCI

95. Support and transferability Software is open source and is available for use by others Regionally scaled versions of the Data Navigator are being adapted for use at the Moss Landing Regional Data Center State Board has expressed interest in adopting the Data Navigator for broader use in association with CEDEN The Council’s Healthy Streams workgroup has expressed interest in adopting the Report Card for broader use in the Healthy Streams web portal

96. ?

97. www.ccamp.org