|
DRAFT 02/19/2014 Exploratory Analysis of STA Monitoring Data Technical Support for Restoration Strategies Science Plan Prepared for U.S. Environmental Protection Agency & U.S. Department of the Interior |
|
Table of Contents
|
|
The information below has been compiled
as a partial basis for evaluating historical and current
designs of the STA monitoring program to support the
Science Plan (questions and projects), as well as
long-term operation/management. The following criteria
and applications are relevant:
The initial focus is on summarizing and
graphing the flow, P load, and concentration data
collected over the past several years (WY 2006-2013).
The displays provide general perspectives on daily,
monthly, and yearly variations in concentration relative
to variations in flow and season. Sampling frequencies
relative to the time scales of major storm events are
also represented. The displays can be used a frame of
reference for future analysis and discussion of the data
and sampling design for individual sites or STAs.
Software used to develop
DMSTA datasets has been
updated and used for data processing and display.
A secondary focus is on evaluating the potential impacts of recent changes in the monitoring program design for STA individual flow paths relative to the historical design, which has been in place since the first STA prototype started operation in 1994 (Everglades Nutrient Removal Project, portion of STA-1W). The design has also been used to measure BMP performance under the EAA Regulatory Rule since ~1993. The historical design (weekly flow-weighted composite and weekly grab samples) is still in place at stations used to formulate the overall mass balance of each STA, as identified in the discharge permits. Auto-samplers have been eliminated from the individual flow paths, except those attached to the permit. Grab sampling frequencies at flow-way starts, interiors, and ends are biweekly, monthly, weekly, respectively. As of April 2013, these changes had apparently not yet been implemented in the STA-2 flow-ways. Sampling at seepage discharge structures has also been eliminated. Depending on magnitude and variability, this could limit the ability to formulate mass balances on some distribution canals and STA cells. This is work in progress. There are no conclusions. The libraries of data summaries and displays provide starting points for future discussion, refinements, interpretations, and applications to support science plan projects and monitoring program optimization.
The data summaries are presented for
future interpretation and discussion. They can be
updated and explored to help support Science Plan
projects, monitoring program design, and modeling.
Software developed in the DMSTA calibration effort has
been refined and applied here.
The links below point to file folders
with charts in various formats. The files are named by
data category & monitoring site. The data were
downloaded in June 2013. Data for some interior sites
have not been compiled. The flow data for some of the
interior STA sites are missing or may not reflect the
most recent calibrations (especially in STA-1E).
Folder Descriptions (Illustrated with
data from the S5A Pump Station)
Other Folders - Currently Not Filled:
Links to File Folders Containing Results
for Various Station Subsets & Time Periods
Daily, Monthly, & Yearly Time Series by Site in PDF Format
Plan to develop interface to access site results. Google Earth or static maps. Example of trends analysis and spatial displays of WCA-3A data.
Elimination of composite samples and
reduction in grab monitoring frequency from weekly to
monthly is simulated by selecting grabs collected on
days 1-7 of each month and deleting the remaining grabs.
Other day ranges (e.g., 8-14, 15-21, etc.) could be
tested, but would not be expected to give significantly
different the long-term average results because the
timing of storm events is not correlated with the day of
the month.
Loads & FWM concentrations computed from
weekly composites using the "standard" technique are
compared with values computed using monthly grabs.
This change was implemented in ~WY 2012 at the interior
cells. To provide a general sensitivity analysis, the
monthly grab frequency is also applied to the other
station types. The biweekly schedule for flow-way
starts is not tested, but would be expected to be within
the range of results for the weekly and monthly
schedules.
Daily, Monthly, & Yearly Time Series Weekly Composites vs. Weekly Grabs - By Site All Sites Weekly Composites vs. Monthly Grabs (Subsampled from Weekly Data) All Sites
Short-term flow/concentration/load
dynamics relative to sampling intervals in selected
years with large storm events (weekly composites &
weekly grabs). Likelihood of capturing concentration
spikes associated with storm events decreases with
decreased grab sampling frequency. Each file contains
two sets of charts (weekly grabs & monthly grabs), as
indicated in the file bookmarks.
Storm event dynamics are also important
for evaluating the algorithm used to construct daily
concentration time series used in DMSTA simulations of
STA performance. The current algorithm is based on
the long-term monthly flow-weighted mean TP
concentrations for each source basin. The algorithm
does not directly account for the fact that
concentrations are typically higher on the rising limb
vs. trailing limb of storm hydrographs ("first-flush
effect"). The same pattern applies to dry-out /
re-wetting events in the STA cells. This limitation
could have significant impacts on simulating the
benefits of Flow Equalization Basins (FEBs), which would
be more likely to capture the rising limbs (before they
fill up).
This analysis evaluates consequences of
eliminating auto-samplers in the individual STA flow
paths, which will be grab-sampled biweekly (starts),
monthly (interior), or weekly (ends). Weekly
auto-samplers will continue to be used at sites need to
establish the overall mass balance (inflows & outflows),
as mandated in the discharge permits. These changes
could have significant impacts on the accuracy and
precision of the mass balances of the individual cells
and flow paths. Further adjustments to the sampling
plan may be needed to provide sufficient data to support
STA research, trend analysis, modeling, and management.
The data summaries are presented for future
interpretation and discussion.
Correlations between yearly flow-weighted-mean TP concentrations computed from weekly grabs, monthly grabs, & weekly flow-weighted composites are summarized below. Monthly grab datasets have been created by selecting weekly grab samples collected on day 1-7 of each month. This would simulate a routine monthly sampling schedule. A biweekly schedule has not been tested, but results are expected to lie between the weekly and monthly results. Results are summarized by STA & Station Type (Inflow, Start, Interior, End, Outflow) for Water years 2006-2012. The data were screened so that at least 3 composite samples were collected in each year. While not explicitly included in the STA monitoring plan, data from major EAA pump stations (S5A, S7, and S8) are also included as representative of flow/concentration dynamics in regional runoff and STA inflows.
FWM concentrations are computed for each
site and water year using three sets of samples:
Three sets of correlations are explored
for each data subset using ln-transformed yearly FWMs:
Each correlation is represented as a
simple ratio (i.e. Y = k X). The value of k is
computed as the ratio of the geometric means across
sites & years in each data subset.
Bookmarks are included in the PDF files to help navigation Ideas for future analysis: Power
for detecting trends a given site depends on random
year-to-year variations, in turn influenced by sampling
frequency and method.
It also seems useful to the evaluate the
impact of sampling frequency & method on the precision
of metrics typically used for tracking P removal
performance (% removal, settling rate, etc).
The recent elimination of auto-samplers
at most flow-way start, interior, and end sites poses a
number of potential problems for tracking and modeling
STA performance. It will be difficult to test and
refine the design model (DMSTA) calibrated to
auto-sampler loads using current and future grab-samples
loads. At most sites, loads historically based
auto-samplers will not be directly comparable to those
currently computed from grab samples. It makes sense
to evaluate and refine the algorithm used to compute
loads from grab samples (linear interpolation) to
minimize differences with auto-sampler values.
That can be done using the historical data when both
grab and composite samples were collected.
Alternative methods used to compute loads from grab samples can be investigated and compared with loads computed from auto-samplers. Estimates of precision in the annual mean loads and FWMs are also desired. Candidates:
Compiling flow and concentration datasets
used for load calculations at individual sites and cell
inflow/outflow boundaries is labor-intensive and
error-prone because of the extensive data manipulations
involved. Comparing my results with SFWMDs could
provide a cross-check. The same applies to water depth
time series. It will also be useful to automate the
process to the extent possible and eventually work from
the same datasets.
Software for load calculations A statistical framework and database developed for the Onondaga Lake monitoring program provides an template for evaluating precision of annual means and power for detecting trends, as related to sampling program design. Precision of Annual Means - STAs
.....
Power for Detecting Trends - STAs
.....
Software for
analysis of trends &
variance components Water & Mass Balances on STA Inflow & Outflow Canals DMSTA Simulations, data thru 2007, compare measured inflows with measured outflows in various canal segments, 30-Day time step, ignoring storage in canal, assuming P is conservative (no net exchange with sediments). DMSTA Application to Track STA Performance Relative to RS Design Assumptions DMSTA has been used to forecast the performance of RS projects, but can also be used as tool for interpreting monitoring data and test the model parameter values used in the RS design simulations. Below are simulations of historical data (thru WY 2011) using STA cell input parameters identical to those used in RS design simulations. Next steps will be to update datasets & re-calibrate model parameters (hydraulics, P uptake, etc.). Yearly Time Series For All STAs 30-Day, Yearly, & Cumulative Time Series for Each STA Inflow Volumes & Concentrations vs. RS Simulations Dashboards for RS Simulation Results Dashboard with Supporting Information & Cell Simulations (under development...) The databases compiled to construct DMSTA calibration datasets can leveraged for other general purposes, such as reporting water & mass balances, tracking trends in performance (%reduction, K, Cout), steady-state modeling, characterization of cell hydraulic properties, etc. Software for automated updating of datasets (daily flow, concentration, inflow stage, outflow stage, rainfall, et..) would facilitate this work.
Hydraulic model calibrations are typically developed
for each cell or flow path.
Parameter estimates can be updated and compared with
cell characteristics (vegetation, topography, shape,
etc). Updated hydraulic parameters can be used in
simulating Restoration Strategies plans, which typically
used default parameter estimates (a=1, b=4).
Correlations between head loss & flow could also
be developed and compared with cell properties.
Restoration Strategies Plan - 04-27-2012
SFWMD Website
Restoration Strategies Science Plan - 06-07-2013
|
| 02/19/2014 http://www.wwwalker.net/ever/sta_monit/index.htm |
