June 1998 Bulletin (Tech.) 905
Assessing Phosphorus Load
Reductions Under Agricultural
Best Management Practices
Ronald W. Rice and Forrest T. Izuno
UNIVERSITY OF
^FLORIDA
Agricultural Experiment Station
Institute of Food and Agricultural Sciences
Ronald W. Rice is Assistant Professor and Crop Nutrition and Water Quality Specialist; Forrest T. Izuno is Professor and Water Management
Specialist; Everglades REC, Belle Glade, FL 334308003.
Assessing Phosphorus Load Reductions
Under Agricultural Best Management Practices
Ronald W. Rice and Forrest T. Izuno
Introduction
The Everglades Forever Act (1994) outlined a comprehensive restoration
program designed to improve water quality and hydroperiod (water quantity and
delivery timing) to the Everglades. Because the Everglades ecosystem evolved as a
lownutrient (oligotrophic) environment, increased deliveries of nutrient enriched waters
from both agricultural and urban sources is an important water quality concern. The
chemical element phosphorus (P) has attracted the most scrutiny since this nutrient is
most likely to encourage undesirable ecosystem changes in aquatic and wetland
ecosystems (Federico et al., 1981; Davis, 1994).
The Everglades Agricultural Area (EAA) is geographically located south of Lake
Okeechobee and north of the Water Conservation Areas (WCAs). The EAA plays an
important role in the Everglades water supply, either directly through agricultural
drainage runoff, or indirectly by serving as a conduit for large water transfers from Lake
Okeechobee to the WCAs. Drainage waters in the EAA can become nutrient enriched
through crop production inputs, natural mineralization (biological breakdown and
nutrient release) of organic "muck" soils, and use of nutrient enriched irrigation waters
I
UNIVERSITY OF FLORIDA LIPPARIES
/o00
F from Lake Okeechobee. Because water storage options are limited by shallow soils
q o with underlying limestone formations that resist percolation, excess water must typically
"SCIENCE be pumped offfarm into area drainage canals serving the EAA. This canal network
LIBRARY
drains a significant portion of EAA runoff into the WCAs through pump stations
managed by the South Florida Water Management District (SFWMD).
Chapter 40E63 of the Everglades Forever Act placed responsibility with the
SFWMD to design the EAA Regulatory Program (SFWMD, 1992). This program
requires that total P (TP) loads in EAA runoff entering the WCAs must be reduced by at
least 25% relative to the 19791988 baseline period of record. This basinlevel target is
to be collectively achieved by all growers through the implementation of best
management practices (BMPs) specifically designed to reduce TP discharge from their
properties. Chapter 40E63 also requires growers to develop their own water quality
monitoring plans consisting of drainage flow measurements and water sample
collection at farm structures discharging into works of the SFWMD (Whalen and
Whalen, 1996).
Basinlevel TP drainage is continuously monitored at seven pump stations
operated by the SFWMD. Recognizing that yeartoyear differences in rainfall
distributions will influence offfarm drainage requirements and overall basin runoff, the
SFWMD developed a model which adjusts EAA basin TP discharge for annual rainfall
distributions; based on comparisons to the 10year period of record for similar data
(SFWMD, 1992). For the first annual compliance determination (May 1995 through
April 1996), the EAA basin recorded a 68% TP reduction (SFWMD, 1996).
2
The EAA Regulatory Program (SFWMD, 1992) is unique in that it assesses
water quality improvements at the basin level rather than at the individual farm level. A
farmscale computer model (EAAMODFARM) currently under development will
eventually provide growers with a sophisticated tool for designing effective Preduction
BMP systems for their farms (Pickering et al., 1997). At present, growers and water
managers need timely feedback regarding the effectiveness of evolving farm BMP
programs.
The objective of this publication is to present several methods that can be used
to quantify TP reductions under BMP implementation at the farm level. Factors that can
complicate or bias the interpretation of farmlevel water quality monitoring data will be
identified. The discussion will emphasize that meaningful water quality interpretations
rely on an adequate assessment of rainfall that occurred during the collection of farm
drainage water quality monitoring data.
Water Quality Monitoring Data
In order to monitor farmlevel TP discharge, the farm discharge points should be
instrumented to automatically collect water samples during drainage events. Some
provision must also be made to track discharge flow (or volume) during the drainage
event. Details regarding water sampling instrumentation and flow determinations,
which generally involve a pump calibration procedure and the automatic or manual
monitoring of canal depth levels, are beyond the scope of this discussion. Suffice it to
3
say that it is important to maintain accuracy and consistency across all monitoring and
data collection protocols.
The EAA Regulatory Program (SFWMD, 1992) calls for the collection of
composite water samples. Thus, over the course of a drainage event, the automatic
water sampler is programmed to obtain an aliquot of water at predefined time intervals
(say every two hours), with all aliquots combined within a single collection vessel. The
TP concentration of this single composite water sample is considered representative
of the drainage event (or events).
Raw Data
Table 1 includes a listing of raw water quality monitoring data collected over a
1month period from a single farm discharge structure. Each data entry is
representative of a 24hr period. During the first four days, there was no pumping
activity (drainage volume=0 gallons) and the TP concentration is simply left blank since
no water sample was collected. The first pumping event, initiated sometime on day 5,
produced a drainage volume of 5,119,240 gallons. Over the course of this event, the
automated water sampling protocol used at this farm site collected a composite sample
with a TP concentration of 0.0882 mg/L.
Table 1 deliberately contains "mixed" units (English and metric). Although
monitoring data are generally collected in English units (gallons), water quality
laboratories often report concentrations in metric units (mg/L). Subsequent use of
these data to generate other calculated data values will invariably require some form of
4
unit conversion. Details regarding units of measure for English and metric systems and
nutrient load calculation protocols for various water sampling strategies are fully
discussed in Izuno and Rice (1998).
Calculated Data: Load
A "load" is defined as a mass of a chemical or chemical compound that is moved
from one location to another. The P load for any given farm drainage event is simply
the mass of P that is present in the given drainage volume, calculated by the product of
the discharge volume with its associated TP concentration value:
load = volume*concentration*{unit conversion factors}. (1)
For the example data in Table 1, the total P load for day 5 is:
load = (5,119,240 gal)*(0.0882 mg/L)*{(3.785 L/gal)*(0.000001 kg/mg)}
= 1.709 kg,
which can be converted to English units:
load = (1.709 kg)*(2.205 Ibs/kg)
= 3.768 Ibs.
Calculated Data: Unit Area Load (UAL)
Nutrient load calculations are biased by farm size. Direct comparisons of P
loads between farms can be misleading since discharge volume is a factor in the load
5
calculation and big farms (with larger watersheds) will typically have greater volumes of
water to drain than smaller farms. To avoid this issue, nutrient loads should be
"normalized" for farm watershed area and reexpressed as "unit area loads" (UALs):
UAL = load/(farm area). (2)
The calculation for day 5 UAL (in English units) is:
UAL = (3.768 lbs)/(2250 acres)
= 0.0017 Ibs/acre.
Although Table 1 lists daily UAL values, it is more common to discuss UAL
values representing longer time periods such as a month or year. Keep in mind that
load calculations should be performed on daily data since the calculation (Equation 1)
addresses specific discharge volumes and concentrations that occur on any given day.
Once the daily load values are calculated, the easiest way to find the monthly UAL
value is to calculate all daily loads (Equation 1), total them to find the monthly load
value (177.003 Ibs; Table 1), and then divide this value by the farm area (Equation 2) to
generate the UAL value for the month (0.0787 Ibs/acre). This value can also be found
by summing up all daily UALs over the month, but this option is more labor intensive.
Calculated Data: Cumulative UAL
The last two data columns in Table 1 present UAL and rainfall values as
cumulative summaries. The cumulative UAL for any given day is simply the total UAL
6
recorded from the first day of monitoring (i=day 1) through the given day of interest
(i=day x):
cumulative UAL (fordyx = E(UAL) for i=day 1 through i=day x. (3)
For the provided example (Table 1), cumulative UAL remains at zero over the first four
days of monitoring when no pumping occurred. On day 5, cumulative UAL (0.0017
Ibs/acre) is exactly equal to the UAL recorded for day 5. Additional pumping activity on
day 6 produced a P load of 35.221 Ibs (Equation 1) which equates to a UAL of 0.0156
Ibs/acre (Equation 2). The cumulative UAL for day 6 is found by applying Equation 3 as
follows:
cumulative UAL (foray) = UAL, + UAL2 + UAL3 + UAL4 + UAL, + UAL6
= 0+0+0+0+0.0017+0.0156
= 0.0173 Ibs/acre.
Note that the cumulative UAL value remains unchanged on days when no
pumping has occurred. Thus, cumulative UALs remain at 0.0412 Ibs/acre from day 10
through day 13 (when UAL=0 Ibs/acre). A UAL of 0.0083 Ibs/acre on day 14 results in
the updated cumulative UAL value of 0.0495 Ibs/acre (Table 1). On day 31, the last
day of the monthly monitoring period, the cumulative UAL value (0.0787 Ibs/acre) is
exactly equal to the monthly UAL total recorded for the 31day period.
Calculated Data: Cumulative Rainfall
Although rainfall is not a direct factor used in load and UAL calculations, an
7
understanding of rainfall and its influence on farm discharge is critically important in the
interpretation of water quality trends. The utility of a cumulative rainfall database will
become apparent in subsequent discussions. For now, it is only necessary to note that
the cumulative rainfall calculation is similar to the cumulative UAL calculation, except
rainfall values from i=day 1 through i=day x are used:
cumulative rainfall (fordayx) = E(rainfall), for i=day 1 through i=day x. (4)
Why Cumulative?
Once the cumulative UAL and rainfall databases have been calculated, there is
no particular need to focus on any given value for any given day. The important aspect
of this exercise is to generate the entire chronological listing of paired daily cumulative
UAL and rainfall values. These paired data are useful for generating graphical
descriptions of incremental UAL discharge over incremental rainfall. The utility of these
graphs for interpreting water quality monitoring data will be discussed shortly.
Factors that Challenge the Interpretation of Monitoring Data
A number of factors can limit the options available for interpreting water quality
monitoring data. It has already been mentioned that P load data should be adjusted for
farm area since volume discharge requirements can be influenced by watershed size.
For example, over similar 15month BMP monitoring periods (Table 2), the total P load
discharged from Farm C (768.2 Ibs) was 60% smaller than for Farm B (1947.9 Ibs).
8
However, after normalizing for farm area, the total UAL discharged from Farm C (1.200
Ibs/acre) was almost three times greater than for Farm B (0.423 Ibs/acre).
Time is also a factor that can bias data interpretations. Calculating the total UAL
discharge for baseline and BMP periods may serve as a convenient data summary, but
these totals should not be compared to one another because the monitoring time
periods differ. For example, the total BMP period UAL (3.251 Ibs/acre) for Farm A is
47% greater than for the baseline period (2.206 Ibs/acre), largely because BMP data
were acquired over an extra 294 days (Table 2).
Normalizing baseline and BMP total UALs by their respective time periods would
minimize the effect of dissimilar monitoring time periods. However, subsequent
comparisons of these UALtotime ratios are not instructive since these ratios fail to
address differences in rainfall distributions (and the influence of rainfall on offfarm
drainage requirements) that existed during baseline and BMP data collection.
An alternative to comparing UALs for entire monitoring periods is to assess UAL
discharge trends over time. A natural inclination is to plot monthly UALs over time but
these graphs generally fail to illustrate trend differences between baseline and BMP
data. For example, some of the highest monthly UALs on record for a research farm
site occurred after the implementation of Preduction BMPs (Fig. la). These
observations highlight the fact that UALs are influenced by rainfall distributions which
vary with season and year (Fig. 1b). In the EAA, water removal through gravity and/or
deep percolation/seepage is negligible due to flat basin topography, shallow soils, and
a relatively impermeable bedrock. Given limited water storage options, farm (and
9
basin) drainage requirements are closely tied to antecedent and current rainfall
conditions.
In late1994, Tropical Storm Gordon caused flooded conditions throughout the
16county SFWMD regional drainage system which limited the efficient drainage of the
EAA basin. During this time period, many farms were operating under BMP plans but,
nonetheless, recorded elevated UAL discharges during a year that developed into the
fifth wettest on record (SFWMD, 1995).
Clearly, UAL data comparisons between monitoring periods must employ
methods that address rainfall profiles that existed during baseline and BMP data
collection. The first inclination is to normalize UAL data for rainfall. Attempts to achieve
a high degree of resolution by normalizing daily (or weekly) UAL data by daily (or
weekly) rainfall totals are not instructive. A problem arises when there is no rainfall
during these short time periods since the UALtorainfall ratio goes to infinity (i.e.
dividing by zero) which is not interpretable. Ironically, normalizing by longer monthly
increments (Fig. 1 c) fails to illustrate clear trends over time because resolution is lost
regarding daily rainfall distributions and their effect on offfarm pumping requirements.
This point is illustrated herein using data from a research farm site. Note that
1994 rainfall totals were similar in November (8.0 inches) and December (8.1 inches)
(Fig. Ib), but monthly UALs (0.83 and 0.36 Ibs/acre, respectively) were strikingly
different (Fig. 1a). No additional information is gained by reviewing normalized monthly
UALtorainfall ratios (Fig. Ic). Inconsistent trends across monthly data values will arise
due to the effects of dissimilar daily rainfall distributions on pumping activities. In this
10
example, November rainfall was delivered as a single 3day tropical storm event and
resulting flooded field conditions demanded unusually high discharges. In December,
rainfall was distributed over three separate events throughout the course of the month.
Under these less extreme rainfall conditions, the grower was able to moderate offfarm
discharge activity.
In summary, a number of constraints limit the options available for interpreting
south Florida water quality monitoring data. The following section highlights water
management trends that have evolved under BMP implementation. Recognizing these
trends is useful in the interpretation of water quality data. This section is followed by a
presentation of three analytical methods that can provide useful assessments of water
quality trends under Preduction BMP technologies.
Water Management Trends Under BMPs
Many growers have improved onfarm water use efficiency in their efforts to
reduce P (i.e. UAL) discharges. In general, crop production operations under BMP
implementation will allow a more conservative water management response to rainfall.
Popular BMP strategies include redesigning water conveyance networks to enhance
farm drainage uniformity and capacity, hydraulically isolating different crop commodities
within contiguous areas to improve water table management, preferentially retaining
highnutrient draindown waters onfarm, using field rather than canal water levels to
schedule irrigation and drainage events, maximizing water removal through
11
evapotranspiration losses, extending fallow flooded field storage periods when
appropriate, and conducting agriculture under higher than traditional water table levels
(Izuno et al., 1995; Bottcher et al., 1995).
Over time, the comprehensive implementation of these strategies (either singly
or in combination) gives growers additional flexibility with respect to water management
decisions. Thus, under BMP operations, improved control over water within the
confines of the farm should lead to reduced offfarm discharge pumping requirements,
particularly in response to minor rainfall events.
Assessing P Reductions Under BMP Implementation
Method 1: Comparing Baseline and BMP UALtoRainfall Ratios
Calculating Meaningful Index Ratios
Given the above trends, an overall summary of water management response
(discharge pumping) to rainfall can be estimated by normalizing monitoring period UAL
totals by their respective rainfall totals (UAL:R). These calculations are performed with
Equations 5 and 6, where i=first day through i=last day of the baseline monitoring
period and j=first day through j=last day of the BMP monitoring period :
baseline UAL:R ratio
= (total baseline period UAL)/(total baseline period rainfall) (5)
or
= E(daily load)/(farm area)]/_(daily rainfall),, and
12
BMP UAL:R ratio
= (total BMP period UAL)/(total BMP period rainfall) (6)
or
= [(daily load)/(farm area)]/E(daily rainfall),.
Using data for Farm B provided in Table 2 (monitoring period total loads and UALs are
given), these equations produce the following UAL:R values:
baseline UAL:R ratio = (1.152 lbs/acre)/(131.0 inches)
= 0.00879 Ibs/acre/inch, and
BMP UAL:R ratio = (0.423 lbs/acre)/(58.5 inches)
= 0.00723 Ibs/acre/inch.
Assessing Ratio Differences
Over time, UAL:R ratios should decline as BMP strategies are phased into crop
production operations. Evidence of a reduction in P discharge can be found when the
UAL:R value for the BMP monitoring period is smaller than for the baseline period. This
comparison can be quantified by finding the relative UAL:R difference, basically
assessing to what extent (as a percentage) the BMP value differs from the baseline
value:
relative UAL:R difference
= [(BMP UAL:R baseline UAL:R)/(baseline UAL:R)]*100. (7)
Using the previously calculated UAL:R values for Farm B, this relative UAL:R difference
is calculated as follows:
13
relative UAL:R difference = [(0.00723 0.00879)/0.00879]*100
= 17.7%.
Thus, with respect to overall baseline and BMP rainfall, the total UAL discharge
during BMP operations at Farm B was roughly 17.7% less than during baseline
operations. A negative value indicates that BMPs are effecting UAL reductions. Similar
calculations for Farm A indicate a slight (+2.7%), but likely negligible, increase in UALs
under BMPs. The +80.1% relative UAL:R difference for Farm C strongly suggests that
BMP strategies are either inadequate or are improperly managed. This methodology
provides rapid feedback regarding progress under BMPs, and allows the timely revision
to the BMP program.
Method 2: Comparing Baseline and BMP Cumulative Data Distributions
Comparisons between monitoring period UAL:R ratios (discussed above) provide
supporting evidence that P reductions are occurring under BMP implementation.
However, these ratios fail to address the influence of daily rainfall distributions on UAL
discharges. Daily resolution can be achieved by reexpressing UAL and rainfall data as
cumulative data (Table 1, Equations 3 and 4). The purpose is to generate a listing of
paired data that reflect UAL and rainfall totals chronologically over time. Descriptions of
farm P discharge profiles are achieved by plotting cumulative UAL as the dependent
variable (Yaxis) and cumulative rainfall as the independent variable (Xaxis). This
description is distinctly advantageous because it assesses water quality trends relative
to rainfall volume rather than calendar time.
14
Using Regression to Compare Cumulative Data Distributions
For comparative purposes, the cumulative values are first compiled separately
for baseline and BMP data and then plotted together. The comparison is quantified
with linear regression applied separately to the baseline and BMP cumulative
databases. For this analysis, regression is intended to serve a descriptive purpose
rather than a predictive function. Differences between baseline and BMP regression
slopes can be quantified by finding the relative slope difference, which assesses to
what extent (as a percentage) the BMP slope value differs from the baseline value:
relative slope difference
= [(BMP slope baseline slope)/(baseline slope)]*100. (8)
For illustrative purposes, Figure 2 presents schematic diagrams of various
outcomes that can be expected with the linear regression exercise (actual cumulative
data scatter plots are not shown). Evidence of consistent P reductions under BMP
operations is found when the BMP slope is of lower magnitude than the baseline slope.
This regression result indicates that the overall BMP data distribution (cumulative UAL
vs. cumulative rainfall) resides below the baseline distribution, which in turn reflects a
consistent decline in UAL discharge in response to rainfall during BMP operations.
Given this situation, the relative slope difference (Equation 8) will produce a negative
value. Although this scenario is depicted in Fig. 2a, the relative slope difference
(Equation 8) is minimal:
15
relative slope difference = [(0.0291 0.0300)/0.0300]*100
= 3.0%.
The reverse scenario (declining water quality trends under BMPs, albeit minor) is
presented in Fig. 2b, whereby the BMP slope is +3.1% larger than the baseline slope.
Differences between baseline and BMP UAL discharge trends are more obvious in Fig.
2c, with a relative slope difference (30.0%) supporting the conclusion that water quality
trends have improved markedly under BMP operations. The reverse scenario (Fig. 2d)
portrays a BMP slope value 42.9% greater than the baseline distribution slope. This
relationship strongly suggests that the BMP program is either not working, is
inadequate in scope, or is improperly implemented.
Application of Statistics to Regression Data
If desired, one can take the regression exercise one step further by assessing
whether or not the difference in baseline and BMP slopes are statistically significant.
The slopes can be compared with a ttest that addresses the entire paired cumulative
data set for both monitoring periods, using the test statistic:
tc = (BMP slope baseline slope)/[(s.e.BMP)2 + (s.e.BL)205, (9)
where s.e.,,P and s.e.BL are the standard errors associated with the regression slope
estimates for the BMP and baseline cumulative distributions, respectively. Given a
"sufficiently large" database, the slopes are significantly different if:
16
Itl > 1.97 ... significant at the 5% level, and (10)
Itjl > 2.60 ... highly significant at the 1% level, (11)
where Itj refers to the absolute value of to. From a statistical standpoint, "sufficiently
large" refers to a combined cumulative baseline and BMP water quality database
exceeding 242 days (this equates to 240 degrees of freedom for the ttest which,
relative to standard statistical tables for the t distribution, minimizes the critical tvalues
to either 1.97 or 2.60, depending on the choice of significance level). In reality, any
meaningful interpretations of water quality data would require monitoring efforts that
greatly exceed 242 days, thus the critical tvalues listed in Equations 10 and 11 are
appropriate.
Examples From Farm Research Sites
The interpretative value of cumulative UAL and rainfall distributions will be
highlighted below for four research farm sites that collectively represent a wide range of
cropping systems and BMP implementation strategies. Figure 3a depicts baseline and
BMP cumulative data distributions for Site UF9206A&B, a mixedcrop (sugarcane,
vegetables, rice, and sod) operation. To facilitate comparisons, linear regression lines
for the baseline and BMP data are also provided. Based solely on linear regression
results, one would conclude that BMPs have failed to improve UAL discharge trends
since the BMP slope value is larger than the baseline slope value. This difference is
quantified with Equation 8, using slope values provided in Table 3:
17
relative slope difference = [(0.04608 0.04302)/0.04302]*100
= +7.1%.
A review of the data scatter plots reveals that new water management practices
during early BMP operations actually supported reduced UAL discharges over the first
46 inches of cumulative rainfall (Fig. 3a). The abrupt UAL increase from 1.21 to 3.68
Ibs/acre occurred during a 2week pumping period in response to an 8inch rainfall
event (Tropical Storm Gordon). Placed in perspective; this 3day storm incurred a
discharge load that exceeded the sum total load incurred over the first 197 days of the
BMP period. As a result, the BMP distribution is displaced above the baseline
distribution (Fig. 3a), masking attempts to reasonably quantify Preduction trends under
BMP operations.
A visual evaluation of the BMP distribution suggests that, once the farm site
recovered from this single aberrant weather event, BMP discharge trends returned to
those existing prior to the storm. Recognizing that south Florida historically encounters
such conditions, the EAA Regulatory Program (SFWMD, 1992) provides for the
exclusion of extreme rainfall conditions from water quality monitoring databases. Basin
wide flooding during this November 1994 storm resulted in the temporary abandonment
of water management BMPs in efforts to drain flooded fields during the sugarcane
harvest and winter vegetable production season. In order to address the
disproportionate effects of this single weather event, baseline and BMP distributions
are presented for cumulative databases with November data omitted (Fig. 3b). An
evaluation of this data subset reveals a consistent Preduction BMP effect of 25.6%
18
(Equation 8) relative to baseline operations. This UAL reduction under BMP operations
is highly significant, confirmed by using regression data (Table 3) to calculate a t, value
of 46.1 (Equation 9), where ItJ=+46.1 which is greater than 2.60 (Equation 11):
t, = (0.03355 0.04512)/[(1.17x104)2 + (2.22x104)20.5
= (0.01157)/(1.37x108 + 4.93x108)05
= (0.01157)/(6.30x10)0.5
= (0.01157)/(2.51x104)
= 46.1.
In contrast to mixedcropping operations, sugarcane monocultures are less
sensitive to immediate water table levels. The ability of sugarcane to withstand short
term flooding and periodic "wet feet" gives growers additional latitude with respect to
discharge pumping. Although cumulative UALs doubled from 0.18 to 0.36 Ib/acre (data
not shown) during Tropical Storm Gordon, farm operations at Site UF9209A (sugarcane
monoculture) quickly recovered and a retum to BMP strategies ultimately supported a
20.2% UAL reduction (Equation 8; Table 3). A BMP effect is clearly evident for this site
when data encompassing the Tropical Storm Gordon period are omitted from the
analysis (Fig. 4). The baseline and BMP cumulative distributions describe a consistent
divergence and the 30.1% difference between slope magnitudes (Equation 8; Table 3)
reflects a consistent attenuation in UAL discharge during BMP operations.
The linear regression exercise may not be universally appropriate to all cropping
operations. Both the baseline and BMP cumulative distributions for Site UF9201A
(Fig. 5), a vegetable monoculture, describe nonlinear UAL discharge profiles. Unlike
other crop production enterprises, pumping activity for vegetable monocultures will
19
periodically be driven by factors unrelated to rainfall. For example, successive cropping
operations throughout the winter production season necessitate scheduled drainage
events (regardless of rainfall) to accommodate field preparation and crop harvest
machinery. In addition, there is no relationship between offfarm discharge and rainfall
during the summer offseason when fields are deliberately maintained under fallow
flooded conditions.
Intermittent plateaus (no change in cumulative UAL over increasing rainfall)
reflect these deliberate fallow flooded field storage periods when rainfall events have no
bearing on drainage requirements (Fig. 5). However, the clear separation between
baseline and BMP distributions support conclusions favoring P reductions under BMP
operations. In particular, note that the BMP distribution begins to diverge rapidly from
the baseline distribution at 15.7 inches of cumulative rainfall, when baseline UALs
shifted vertically from 4.36 to 5.87 Ibs/acre. This UAL increase under zero rainfall
conditions during baseline operations represents the deliberate discharge of summer
fallow flood waters to accommodate September planting schedules. Under BMP
operations, modified water management strategies and new drainage practices were
implemented to address this traditional "September UAL spike" period. The diverging
distributions (Fig. 5) document declining UAL discharge trends under these BMP
strategies.
As ar alternative to regression for these type of data, one can take the baseline
and BMP cumulative UAL values recorded at (for this example) every 10inch rainfall
increment through 70 cumulative inches (the limit of the BMP distribution; Fig. 5). Using
20
these values, an average cumulative UAL can be calculated for the BMP (4.69 Ibs/acre)
and baseline periods (7.48 Ibs/acre). These average values can subsequently be used
in place of slopes (Equation 8) to calculate an overall 37.3% UAL reduction under BMP
operations.
As schematically described in Fig. 2d, comparisons of cumulative data
distributions will also describe declining water quality trends. This scenario is illustrated
for Site UF9204A (Fig. 6) whereby the BMP cumulative distribution remains consistently
above the baseline distribution. This site was under sugarcane production during
baseline operations but five months into BMP operations, half the farm area was
rotated into flooded rice. This major cropping modification occurred without the
concurrent implementation of an adequate water management plan. The 130.6% UAL
increase during the BMP period (Table 3) reflects the inability to properly manage rice
drainage waters at this site. It is important to emphasize that rice production is a
perfectly feasible production practice, given the implementation of appropriately
designed BMPs. The adoption of improved management strategies at a different
sugarcanerice farm (data not shown) supported a 25.7% UAL reduction during BMP
operations.
Method 3: Comparing RainfallAdjusted UALs
The EAA Regulatory Program (SFWMD, 1992) requires the EAA basin to
achieve a minimum 25% P load reduction. All growers in the EAA are collectively in
compliance if the basin target is met. The annual compliance evaluation involves a
21
model that adjusts annual EAA subbasin P loads for hydrologic variability (SFWMD,
1992). The model also allows for the calculation of estimated farmlevel rainfall
adjusted UALs (AUALs). The model calculates a rainfalladjusted load based on the
comparison of current "water year" load (and monthly rainfall distribution) to an earlier
10year (19791988) period of record database. By definition, the 1994 water year
(WY94) is the 12month period beginning May 1, 1993 and ending April 30, 1994.
Required Data for AUA'L Calculations
In order to calculate farmlevel AUAL estimates for a given WY, the following
information and data are required:
1. Identify the farm location with respect to EAA subbasin (S5A, S6, S7, or S8),
2. Identify the water year of interest,
3. Calculate the farm's total UAL for the given WY,
4. Obtain the Thiessenweighted monthly rainfall values specific to the WY and
EAA subbasin, and
5. Obtain the "fixed coefficients" (rainfall variation, skewness, and adjustment
factors) specific to the EAA subbasin that are used in the model equation.
At this writing, the Thiessenweighted rainfall values and fixed coefficients were
available from the SFWMD website at .
Among several available files, download the archived "EAAB9709.zip" file, "unzip" this
file into its various components, and then open the file titled "EAABASIN.WK4" (a
Windows operating system Lotus 123 file). The EAABASIN.WK4 file has multiple
pages. For Thiessenweighted monthly rainfall values, select the page tab titled
"Monthly" to access the data table titled "Basin Compliance Calculations EAA
22
Regulatory Rule", scroll to the table subheading titled "Monthly Calculated Values:
Basin Average Rainfall (inches)" which includes Thiessenweighted rainfall values
(beginning October 1978) for EAA subbasins S5A, S6, S7, and S8. To obtain the fixed
coefficients for each EAA subbasin, select the page tab titled "Coefficients" to access
the data table titled "Fixed Coefficients for EAA Basin Calculations (From Rule Text)",
and scroll to the table subheading titled "od Mean Rainfall Statistics".
AUAL Calculation
An example of an AUAL calculation is provided below using monthly load data
collected during WY96 from a research farm site located in the S5A subbasin. These
data are provided in Table 4a along with a listing of S5A subbasin Thiessenweighted
monthly rainfall values for WY96. For the readers convenience, the fixed coefficients
for all four subbasins are provided as well (Table 4b).
In order to calculate the estimated farm AUAL, eight sequential calculations must
be performed (SFWMD, 1992). First, calculate three different factors (ml, m2, and m3)
specific to the Thiessenweighted subbasin rainfall values, where ri=individual monthly
rainfall values from i=1 (May) to i=12 (April). Note that m, is simply the average monthly
rainfall for the WY in question. The factor m2 is the average of the squared differences
between individual monthly rainfall values and the average WY rainfall value, while m3
is the average of the cubed deviations:
m, = E(r,)/12, (12)
m2 = ,(r m)2/12, and (13)
23
m3 = F(r, mi)3/12. (14)
Second, calculate the "predictors" (X, C, and S) from the first three rainfall factors. Note
that X is actually the natural log (In) of the total WY rainfall, C is the standard deviation
of the monthly rainfall values divided by the average monthly rainfall value (also known
as the coefficient of variation), and S is a rainfall skewness coefficient:
X = In(12*ml), (15)
C = [(12/11)*m2]S/mi, and (16)
S = [(12/11)*m3]/(m2)1.5 (17)
Third, calculate the adjusted subbasin rainfall value (Ra) for the WY using the
"predictors" and the fixed coefficients (Cm and S,) specific to subbasin of interest
(Table 4b), where "exp" signifies the inverse of the natural log function:
Ra = exp[X + 1.053*(CCm) 0.1170*(S Sm)]. (18)
Finally, calculate the adjusted unit area load for the WY, using the adjusted subbasin
rainfall value (Ra) and the fixed rainfall coefficient (Ram) specific to the subbasin of
interest (Table 4b):
AUAL = UAL*(R./Ra)2868. (19)
Using the data provided in Tables 4a and 4b, these equations produce the following
calculated values to arrive at a WY96 AUAL estimate of 0.5112 Ibs/acre:
24
m, = (1.33 + 6.56 + ... + 1.73)/12
= 4.4900,
m2 = [(1.33 4.49)2 + (6.56 4.49)2 + ... + (1.73 4.49)]/12
= 14.1591,
m3 = [(1.33 4.49)3 + (6.56 4.49)3 + ... + (1.73 4.49)3]/12
= 39.4395,
X = In(12*4.49)
= 3.9868,
C = [(12/11)*14.1591]05/4.49
= 3.9302/4.49
= 0.8753,
S = [(12/11)*39.4395]/(14.1591)15
= 43.0249/53.2787
= 0.8075,
Ra = exp[3.9868 + 1.053*(0.8753 0.7636) 0.1170*(0.8075 0.9999)]
= exp(4.1269)
= 61.9855, and
AUAL = 0.9319*(50.31/61.9855)2.868
= 0.9319*0.5496
= 0.5122 Ibs/acre.
The above AUAL calculation is specific to WY96 for a farm site located in the
S5A subbasin. The procedure can be used to calculate AUAL estimates for any farm
and WY period, given the use of monthly Thiessenweighted rainfall values and farm
UAL data specific to the WY of interest. For any given farm location within a given sub
basin, the same fixed coefficients (Cm, Sm, and Ram; Table 4b) are used for any WY
period. Keep in mind that baseline and BMP monitoring periods for any given farm site
will not coincide with the WY calendar period. In general, water quality monitoring data
25
collected from EAA farms during WY94 will predominantly reflect baseline practices,
WY95 data will represent the transition from baseline to BMP operations, and WY96
data will exclusively reflect BMP operating conditions. For any given farm, evidence of
Preductions under BMPs will be found with lower AUAL values for WY95 and WY96
relative to WY94.
These comparisons can be quantified by finding the relative AUAL difference
across any two WYs of interest (Equation 20), where WY, is the most recent WY and
WYj is the oldest WY in the given comparison:
relative AUAL difference
= [(AUALw, AUALwYj)/(AUALwYj)]*100. (20)
Table 4c provides a summary of UAL and AUAL values for three different WY. The
WY95 to WY94 comparison of AUALs is calculated as follows:
relative AUAL difference
= [(0.9735 1.9502)/(1.9502)]*100
= 50.1%.
Thus, the 12month AUAL incurred during the transition from baseline to BMP
operations (WY95) represents a 50.1% reduction in P discharge relative to the AUAL
discharge under baseline operations (WY94). Under full BMP implementation (WY96
AUAL value of 0.5122 Ibs/acre), a reduction of 73.7% was realized (Table 4c).
26
Summary
With less than four years of monitoring data, researchers and water managers
are challenged to assess the effectiveness of farmlevel Preduction BMPs in south
Florida. Straight comparisons between baseline and BMP period UAL data are
inadequate because calendar time frames and rainfall distributions differ for the two
monitoring periods. Meaningful comparisons require some measure of hydrologic
adjustment to UAL data.
Three methods for comparing water quality monitoring data were discussed.
The first method minimizes rainfall differences (which influence drainage pumping)
across monitoring periods by comparing baseline period total UAL to rainfall (UAL:R)
ratios to those for the BMP period. The second method involved the reexpression of
water quality databases into cumulative UAL and cumulative rainfall values. These
cumulative databases are plotted to assess differences in baseline and BMP UAL
discharge trends over incremental rainfall. Linear regression applied to these
distributions allows differences to be quantified through slope comparisons. Finally, the
application of a hydrologic model (developed for P discharge regulatory compliance
assessments) to farm UAL data allows the calculation of rainfalladjusted UALs
(AUALs) for different water year (WY) periods. Subsequent AUAL comparisons across
different WYs serves to quantify P discharge trends over time.
Although not specifically addressed in this publication, it may be instructive to
briefly summarize water quality trends recorded at 10 EAA research farm sites from
27
late1992 through April 1996 (Izuno and Rice, 1997). Using the UAL:R ratio
comparison method, BMP data for six of 10 sites reflected P reductions of 3 to 33%.
Using cumulative databases, the BMP distribution slope magnitudes for six sites were 6
to 35% lower than for baseline, evidence of longterm P load reductions under BMP
strategies. Eight of 10 sites reflected reductions after omitting nonrepresentative UAL
data collected under flooded conditions caused by a 3day tropical storm. Applying the
hydrologic adjustment model, average AUALs for eight sites declined by 73% over a 3
year period. Across all three analytical exercises, two sites consistently demonstrated
declining water quality trends as a consequence of large cropping system modifications
in the absence of adequate hydraulic BMP technologies. Despite short baseline
monitoring periods and less than four years of data collected under conditions of highly
variable rainfall, analytical methods discussed herein consistently verify BMP reductions
for a wide range of agricultural cropping systems.
28
References
Bottcher, A.B., F.T. Izuno, and E.A. Hanlon. 1995. Procedural guide for the
development of farmlevel best management practices for phosphorus control in
the Everglades Agricultural Area. Univ. of Florida Cooperative Extension Service
Circular No. 1177. Gainesville, Fla.
Davis, S.M. 1994. Phosphorus inputs and vegetation sensitivity in the Everglades. In
Everglades: The Ecosystem and Its Restoration, S.M Davis and J.C. Ogden
(eds), ch. 15, 357378. St. Lucie Press, Delray Beach, Fla.
Everglades Forever Act. 1994. Amendment of the 1991 Marjory Stoneman Douglas
Everglades Protection Act. Chapter 373.4592, Florida Statutes. Tallahassee, Fla.
Federico, A.C., F.E. Davis, K.G. Dickson, and C.R. Kratzer. 1981. Lake Okeechobee
water quality studies and eutrophication assessment. South Florida Water
Management District Technical Publication No. 812. West Palm Beach, Fla.
Izuno, F.T., and R.W. Rice. 1998. Calculating nutrient loads. Univ. of Florida
Cooperative Extension Service Circular No. Gainesville, Fla (in review).
Izuno, F.T., and R.W. Rice (eds). 1997. Implementation and verification of BMPs for
reducing P loading in the EAA. Phase V Final Report (vol. 1) submitted to the
EAA Environmental Protection District. Belle Glade, Fla.
Izuno, F.T., A.B. Bottcher, F.J. Coale, C.A. Sanchez, and D.B. Jones. 1995. Agricultural z
BMPs for phosphorus reduction in South Florida. Transactions of the ASAE.
38:735744.
Pickering, N.B., A.B. Bottcher, and J.D. Stuck. 1997. EAAMODFARM: Everglades
agricultural area farmscale hydrologic and phosphorus transport model version
2.0. In F.T. Izuno and R.W. Rice (eds). Implementation and verification of BMPs
for reducing P loading in the EAA. Phase V Annual Report (vol. 2) submitted to
the EAA Environmental Protection District. Belle Glade, Fla.
SFWMD. 1992. Rules of the South Florida Water Management District Everglades
Regulatory Program. Chapter 40E63, Florida Annotated Code. New: Jan. 22,
1992; Amended: July 7, 1992. South Florida Water Management District, West
Palm Beach, Fla.
29
SFWMD. 1995. Managing a very wet year. South Florida Water Management District,
West Palm Beach, Fla.
SFWMD. 1996. Rule 40E63 information update: EAA basin total phosphorus levels
(June 24). South Florida Water Management District, West Palm Beach, Fla.
Whalen, P.J., and B.M. Whalen. 1996. Nonpoint source best management practices
program for the Everglades Agricultural Area. ASAE Paper No. 962071. St.
Joseph, Mich.: ASAE.
30
Table 1. Example of farmlevel water quality monitoring data.
Raw data Calculated data
Day Rainfall Drainage Total P P P UALT Cumulative Cumulative
volume concentration load load UAL rainfall
inches gallons mg/L kg Ibs Ibs/acre Ibs/acre inches
1 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0
5 2.23 5119240 0.0882 1.709 3.768 0.0017 0.0017 2.23
6 1.32 47847160 0.0882 15.973 35.221 0.0156 0.0173 3.55
7 0.08 43287950 0.0839 13.747 30.311 0.0135 0.0308 3.63
8 0.43 25141490 0.0796 7.573 16.698 0.0074 0.0382 4.06
9 0 10248115 0.0796 3.087 6.806 0.0030 0.0412 4.06
10 0 0 0 0 0 0.0412 4.06
11 0 0 0 0 0 0.0412 4.06
12 0 0 0 0 0 0.0412 4.06
13 1.21 0 0 0 0 0.0412 5.27
14 0.19 22917153 0.0974 8.449 18.629 0.0083 0.0495 5.46
15 1.27 6505400 0.0811 1.997 4.403 0.0020 0.0515 6.73
16 0.15 32179070 0.0648 7.892 17.403 0.0077 0.0592 6.88
17 0.28 20449179 0.0648 5.016 11.059 0.0049 0.0641 7.16
18 0 12774840 0.0613 2.964 6.536 0.0029 0.0670 7.16
19 0 6309890 0.0578 1.380 3.044 0.0014 0.0684 7.16
20 0 0 0 0 0 0.0684 7.16
21 0 0 0 0 0 0.0684 7.16
22 0.29 0 0 0 0 0.0684 7.45
23 0 0 0 0 0 0.0684 7.45
24 0.75 0 0 0 0 0.0684 8.20
25 0 21791085 0.0848 6.994 15.422 0.0069 0.0752 8.20
26 0 0 0 0 0 0.0752 8.20
27 0 0 0 0 0 0.0752 8.20
28 0.44 0 0 0 0 0.0752 8.64
29 0 11845960 0.0779 3.493 7.702 0.0034 0.0787 8.64
30 0 0 0 0 0 0.0787 8.64
31 0 0 0 0 0 0.0787 8.64
Total 8.64 266416532 n.a. 80.274 177.003 0.0787 n.a. n.a.
t UAL = unit area load (assume farm area = 2250 acres).
Table 2. Example of calculating and comparing unit area load (UAL) to rainfall ratios (UAL:R) for baseline and
BMP water quality monitoring data.
Site Farm Monitoring Calendar Time Rain P UAL UAL:R Relative UAL:R
name area period period load ratio differences
acres days inches Ibs Ibs/acre Ibs/acre/inch %
Farm A 1280 baseline 07/23/92 12/31/93 527 85.4 2823.1 2.206 0.02583
BMP 01/01/94 03/31/96 821 122.6 4161.0 3.251 0.02652 + 2.7
Farm B 4608 baseline 07/24/92 12/31/94 891 131.0 5308.1 1.152 0.00879
BMP 01/01/95 03/31/96 456 58.5 1947.9 0.423 0.00723 17.7
Farm C 640 baseline 07/15/92 12/31/94 900 159.0 1029.0 1.608 0.01011
BMP 01/01/95 03/31/96 456 65.9 768.2 1.200 0.01821 + 80.1
t A negative "relative UAL:R difference" indicates that the BMP period UAL:R ratio is smaller than the
baseline UAL:R ratio by the given percentage.
Table 3. Linear regression statistics for baseline and BMP cumulative unit area load (UAL) vs. cumulative
rainfall distributions for four research farm sites.
Baseline period BMP period BMP to baseline comparison
Site rt Slope s.e.t r Slope s.e. Itel Relative slope
name differences
Ibs/acre/inch Ibs/acre/inch %
All monitoring data
UF9201A # 0.925 0.15489 1.44x103 0.916 0.12421 1.28x103 15.9 19.8
UF9204A 0.978 0.00859 2.74x105 0.993 0.01970 6.91x105 149.5 +129.3
UF9206A&B 0.971 0.04302 2.19x104 0.982 0.04608 1.88x104 10.6 + 7.1
UF9209A 0.997 0.01011 1.65x10"5 0.988 0.00807 2.31x105 71.9 20.2
November monitoring data omitted
UF9201A 0.921 0.15228 1.52x103 0.931 0.10835 1.22x10 3 22.5 28.8
UF9204A 0.984 0.00859 2.64x105 0.993 0.01981 7.75x105 137.0 + 130.6
UF9206A&B 0.974 0.04512 2.22x104 0.995 0.03355 1.17x104 46.1 25.6
UF9209A 0.996 0.01007 1.90x105 0.995 0.00704 1.39x105 128.7 30.1
t r = correlation coefficient.
$ s.e. = standard error (square root of the variance) of the regression for the slope estimate.
t' = test statistic for comparing baseline and BMP regression slopes, which differ significantly at the 5%
level when Itl > 1.97 and at the 1% level when Itl > 2.60.
A negative "relative slope difference" indicates that the BMP slope value is of lower magnitude (by the
given percentage) than the baseline slope value.
# Strict application of linear relationship estimates are not recommended for UF9201A data (see Fig. 5).
Table 4a. Example data used in rainfalladjusted unit area load (AUAL)
calculation for the 1996 water year (WY96).
Month Calendar Subbasin Farm Farm
year rainfall P load UALt
inches Ibs Ibs/acre
May 1995 1.33 0 0
June 1995 6.56 119.60 0.0934
July 1995 6.77 90.50 0.0707
August 1995 10.32 232.37 0.1815
September 1995 4.12 97.94 0.0765
October 1995 12.22 287.03 0.2242
November 1995 1.22 0 0
December 1995 0.81 0 0
January 1996 1.96 193.56 0.1512
February 1996 0.59 26.46 0.0207
March 1996 6.25 91.65 0.0716
April 1996 1.73 53.74 0.0420
1996 WY total 53.88 1192.85 0.9319
Monthly average 4.49 99.40 0.0777
t Thiessenweighted rainfall values for EAA S5A subbasin.
t UAL = unit area load (assume farm area = 1280 acres).
Shaded values are used in the WY96 AUAL calculation example.
Table 4b. Fixed coefficients for EAA subbasins.
Subbasin C, Sm Ram
S5A 0.7636 0.9999 50.31
56 0.7302 0.7476 49.77
S7 0.7198 0.6112 4627
S8 0.7821 0.8409 45.68
Shaded values are used in the WY96 AUAL calculation example.
Table 4c. Example of AUAL data comparisons for three consecutive WY.
UAL AUAL Relative AUAL difference from:
WY94 WY95 WY96 WY94 WY95 WY96 WY95 to WY94 WY96 to WY94
bs/acre/year  % change 
1.4461 2.1734 0.9319 1.9502 0.9735 0.5122 50.1 73.7
1.0
a)
0 .8 .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0.8
o .... 6  
0.4
0.0
ASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMA
b)
12
1 2 8 ............................................................. ........................
.C
'4
4ll   I I I~ ll  1i 
  1 Ii1 I L
ASONDJFMAMJJASONDJFMAMJJASONDJFMAMJJASONDJFMA
0.25
C)
0.20  
o
2 0.15 ..
0.00 
0.00    J    ^    
ASONDJ FMAMJ JASONDJ FMAMJ JASONDJ FMAMJ JASONDJ FMA
Month
1992 1993 1994 1995 1996
Baseline monitoring period BMP monitoring period
Fig. 1. Example of monthly a) unit area load (UAL), b) rainfall, and c) UAL to rainfall (UAL:R) ratio
data for baseline and BMP monitoring periods.
a) b)
o BL BL
BMP BMP
0 BL slope = 0.0300 Ibs/acre/inch BL slope = 0.0291 lbs/acre/inch
BMP slope = 0.0291 Ibs/acrelinch BMP slope = 0.0300 Ibs/acre/inch
Difference = 3.0% Difference = + 3.1%
c) d)
S1
:2
E
O BL slope = 0.0300 Ibs/acre/inch BL slope = 0.0210 Ibs/acre/inch
BMP slope = 0.0210 Ibs/acre/inch BMP slope = 0.0300 Ibs/acre/inch
Difference = 30.0% Difference = +42.9%
Cumulative Rainfall (inches) Cumulative Rainfall (inches)
Fig. 2. Schematics of linear regression relationships for baseline (BL) and BMP cumulative databases
depicting a) negligible differences, b) negligible differences, c) improved UAL trends under BMPs,
and d) declining UAL trends under BMPs.
6
a)
46 ______
5     
4
Cumulative Rainfall (inches)
6
b)
4 . . . . .. . . . . . . . .. . . . . ...       .  .             .
^ ^ > ^ ^
4 ?  .i i
1 ..  ... ...... .. BL BL V BMP BMP 
0 20 40 60 80 100 120 140 160
Cumulative Rainfall (inches)
Fig. 3. Site UF9206A&B baseline (BL) and BMP monitoring period cumulative rainfall versus cumulative
unit area P load (UAL) distributions for a) all monitoring data, and b) a data subset omitting the
disproportionate impact of a single tropical storm event.
1.0
0 .8                                          
0.4
00.2  a BL BL v BMP BMP
0.0
0 20 40 60 80 100 120 140 160
Cumulative Rainfall (inches)
Fig. 4. Site UF9209A baseline (BL) and BMP monitoring period cumulative rainfall versus cumulative
unit area P load (UAL) distributions for a data subset omitting the disproportionate impact of a
single tropical storm event.
12
10
10 
8    
j
j56      
M
2 . .................. BL  BL V BMP BMP
0 20 40 60 80 100 120 140 160
Cumulative Rainfall (inches)
Fig. 5. Site UF9201A baseline (BL) and BMP monitoring period cumulative rainfall versus cumulative
unit area P load (UAL) distributions for a data subset omitting the disproportionate impact of a
single tropical storm event.
1.5
*S 0.6 =  . a _ . . ..  .
70
0.6
E
r 0.3 . BL BL BMP BMP 
0.0
0 20 40 60 80 100 120 140 160
Cumulative Rainfall (inches)
Fig. 6. Site UF9204A baseline (BL) and BMP monitoring period cumulative rainfall versus cumulative
unit area P load (UAL) distributions for a data subset omitting the disproportionate impact of a
single tropical storm event.
