In Reply Refer To: July 13, 1992 Mail Stop 412 OFFICE OF WATER QUALITY TECHNICAL MEMORANDUM 92.10 Subject: PROGRAMS AND PLANS--Phosphorus Methods and the Quality of Phosphorus Data PURPOSE In 1990 and 1991, two changes were implemented at the National Water Quality Laboratory (NWQL) for the analysis of phosphorus. Both changes improved the quality of data produced, with resulting implications for the use and interpretation of results. The purposes of this memorandum are to: 1. Document the chronology of phosphorus methods used at the NWQL; 2. Assess the magnitude of bias produced by the historic phosphorus methods; 3. Discuss the implications for data interpretation of the bias in historic phosphorus data; and 4. State the process for rigorous documentation of new methods at the NWQL and inclusion of field quality control (QC) samples in water-quality programs and projects. CHRONOLOGY Since 1973, the chronology of phosphorus methods at the NWQL has been: __________________________________________________________________ Digestion Measurement Period Method Step Step 1973 - I-2600/I-4600 potassium orthophosphate April 30, without a persulfate and by phosphoantimonyl- 1990 dilution SOP sulfuric acid; molybdenum blue 30 min. at 121oC (mixed reagent in an autoclave procedure) Murphy at 15-20 psi & Riley (1962) May 1, I-2600/I-4600 same as above same as above 1990 - with a dilution Sept. 30, SOP 1991 Since I 2610/I-4610 Kjeldahl reagents orthophosphate Oct. 1, (sulfuric acid by phosphoantimonyl- 1991 and potassium molybdenum blue sulfate); 15 min. (separate reagent at 375oC in a procedure) Murphy Kjeldahl block & Riley (1962) digester __________________________________________________________________ U.S. Geological Survey TWRI method I-2600/I-4600 was initially developed based on U.S. Environmental Protection Agency (U.S. EPA) method 365.1. For the digestion step, both methods rely on oxidation of polyphosphate and organic phosphorus compounds to orthophosphate by persulfate ions and sulfuric acid. During digestion, the conversion of polyphosphates occurs more readily than conversion of organic phosphorus compounds that occur in samples either as part of organic particles or attached to mineral particles. Apparently, sometime after 1973, during a method rewrite, a miscalculation or typographical error was introduced into method I-2600/I-4600. The error resulted in the use of low concen- trations of persulfate and sulfuric acid for the digestion step. The error went unnoticed until 1989, when the Ohio District began splitting samples and submitting aliquots to both the NWQL and the Heidelberg College Water-Quality Laboratory (HCWQL). For certain samples, the NWQL method resulted in lower total phosphorus concentrations than the HCWQL method. The differences were especially notable on samples with suspended-sediment concentrations exceeding 50 mg/L. During early 1990, studies at the NWQL confirmed that method I-2600/I-4600 caused reagent-limiting, incomplete digestions (total phosphorus concentrations were therefore biased low) on samples having high concentrations of suspended sediment. It was not determined whether the reagent limitation resulted from high concentrations of suspended sediment, high concentrations of associated organic carbon, or both. The studies did indicate that the reagent limitation could be overcome for most high phosphorus samples if "off scale" samples--those found to have total phosphorus concentrations exceeding 1.0 mg/L--were diluted tenfold (1 + 9) prior to redetermination. Previously, the extent of dilution for off scale samples was discretionary. Accordingly, on May 1, 1990, an SOP for tenfold dilution of off scale samples (>1.0 mg/L total phosphorus) was implemented for method I-2600/I-4600. The SOP also required that all samples appearing "--particularly colored or sediment laden are diluted by a factor of 2-5 fold at the analyst's discretion--" prior to digestion and analysis. Method I-2600/I-4600 with the added SOP was the official method for phosphorus analysis at the NWQL for the period May 1, 1990 through September 30, 1991. Beginning October 1, 1991, a Kjeldahl digestion method--similar to U.S. EPA method 365.4--as modified by Jirka and others (1976) and Bowman and Delfino (1982)--was implemented at the NWQL for both phosphorus and nitrogen. The new phosphorus method (I-2610/I-4610) completely replaced method I-2600/I-4600. The digestion step in method I-2610/I-4610 involves very rigorous conversion--through reduction and hydrolysis--of polyphosphates and organic phosphorus compounds to orthophosphate using sulfuric acid and potassium sulfate at 375oC. Because of the very rigorous digestion, method I-2610/I-4610 avoids either a reagent limitation or an interference with reagents by samples having high concentrations of suspended sediment, organic carbon, or associated constituents. For these reasons, no dilution step is required for analysis of phosphorus in any type of sample. Method I-2610/I-4610 was implemented at the NWQL because it: (a) improved data quality, and (b) markedly reduced sample handling and analyst time. MAGNITUDE OF BIAS IN PHOSPHORUS DATA CAUSED BY METHOD I-2600/I-4600 AT SELECTED SITES IN ILLINOIS The best information for assessing the possible effect of the pre-May 1, 1990, application of method I-2600/I-4600 on USGS phosphorus measurements stems from samples collected at seven Illinois stream sites from 1986 through 1989. A total of 147 samples were collected at the Illinois sites during the period. All samples were collected by USGS personnel, split, and then analyzed for total and dissolved phosphorus by both the Illinois EPA (IEPA) and the USGS. All IEPA measurements used the U.S. EPA Persulfate Method 365.1 (the method on which USGS I-2600/I-4600 was originally based). Analyses of suspended sediment and total organic carbon (TOC) were also available for 126 and 77 of the samples, respectively. Differences in total phosphorus concentrations by the two methods (computed as IEPA measurement minus USGS measurement) are examined here for: (a) several concentration ranges of total phosphorus, suspended sediment, and TOC (Table 1), and (b) geographic (station) location (Table 2). In addition, differences in dissolved phosphorus concentrations are compared for selected ranges of concentrations (Table 3). Each comparison is based on analysis of the comparative data by the Wilcoxon Signed Rank Test--a nonparametric procedure for determining if medians of two populations are statistically different. For this memorandum, a probability (p) of 2 0.05 was selected as representing a statistically significant difference. In general, USGS total phosphorus values are lower than those of IEPA with the largest and most statistically significant differences in samples having high concentrations of total phosphorus, suspended sediment, and TOC (Table 1). For the full total phosphorus concentration range of 0.05 to 2.0 mg/L, a highly significant (p <0.001) median difference of 0.01 mg/L is detected. For samples with relatively high phosphorus concentrations (above 0.2 mg/L), highly significant differences ranging from 0.01 to 0.03 mg/L are detected for the concentration ranges examined. At concentrations below 0.2 mg/L, the median difference of 0.01 mg/L is not statistically significant. Larger negative bias is observed in USGS total phosphorus measurements for samples with high concentrations of suspended sediment. Statistically significant median differences of 0.04 and 0.06 mg/L occur for samples with suspended sediment concentrations above 75 mg/L and 100 mg/L, respectively. For samples with high concentrations of both total phosphorus (>0.2 mg/L) and suspended sediment (>100 mg/L), statistically significant median differences of 0.09 and 0.18 mg/L are observed. Although fewer TOC measurements are available for statistical analysis, the negative bias in USGS total phosphorus measurements again is larger for samples with high TOC concentrations. Although the results show some ambiguity, the effect of varying quantities of suspended sediment and TOC on the measurement bias of USGS total phosphorus concentrations is also seen in Table 2. Statistically significant median differences in total phosphorus are observed for the Little Wabash River, Rock River, and Spoon River sites, which have high median concentrations of suspended sediment, together with high percentages of particulate phase phosphorus (low DP/TP ratios). The two stations predominantly influenced by point sources and having moderate to low suspended sediment concentrations--the Illinois River and Sangamon River sites--display no significant bias and have an estimated median difference of zero. These sites have relatively high concentrations of total phosphorus, but nearly 75 percent is in the dissolved phase. Table 3 compares results for dissolved phosphorus. Although estimated median differences between IEPA and USGS data range from -0.02 to 0.01, these differences are not statistically significant (p 2 0.05) for any of the concentration ranges examined. We believe the lack of statistically significant bias in the USGS dissolved phosphorus data (in contrast to the total phosphorus data) results from a relative lack of suspended sediment and organic carbon in the analyzed filtrates. MAGNITUDE OF BIAS IN PHOSPHORUS DATA PRODUCED BY METHOD I-2610/I-4610 AS COMPARED TO METHOD I-2600/I-4600 INCORPORATING THE DILUTION SOP Charles Patton and Earl Truitt of the NWQL developed method I-2610/I-4610 for phosphorus and Kjeldahl nitrogen. As part of the method proveout, duplicate analyses were run on 1,572 samples to compare phosphorus results from this new method to results from method I-2600/I-4600 with the addition of the sample dilution SOP (three- to fivefold for colored and/or sediment laden samples; tenfold for off scale samples). The first set of duplicates was run on the 417 samples received at the NWQL during April 1991 for which Districts requested both nitrogen and phosphorus analyses. Because of District concerns about the change to a new method, another set of duplicates was run on 1,155 samples received at the NWQL during the period July-September 1991. This set included: (a) the approximately 100 NASQAN samples that arrived for analyses during the period, and (b) about 1,050 samples for which Districts specifically requested phosphorus to be analyzed by both the new and old methods. Prior to statistical analysis, both the April and July-September data sets were sorted by sample type into four groups: unfiltered and filtered surface water; and unfiltered and filtered ground water. Each group of data produced by the two analytical methods was then compared using: (a) all samples within each group, and (b) specific concentration ranges of phosphorus within each group. The comparisons were again based on the Wilcoxon Signed Rank Test to determine if medians of the two populations are statistically different. The test results--shown in Tables 4 and 5--compare the median differences found by subtracting estimated medians of phosphorus measurements by method I-2600/I-4600 from estimated medians of measurements by method I-2610/I-4610. Thus, a positive bias means that the new method produced higher concentrations than the old method. For surface water (Table 4), median phosphorus concentration for the inclusive concentration ranges showed a positive bias during both time periods of 0.02 milligrams per liter (mg/L) for unfiltered samples and about 0.01 mg/L for filtered samples. For the discrete concentration ranges tested, the positive bias increased progressively with phosphorus concentration for both the unfiltered and filtered samples in each time period. For example, for the unfiltered samples during July-September, the bias was 0.003 for the 0.00-0.15 mg/L range, 0.03 for the 0.15-0.30 mg/L range, and 0.3 mg/L for the 2.00-4.50 mg/L range. The bias was statistically significant in 22 of the 24 cases. The two exceptions were for filtered (dissolved) samples: (a) the 0.22-0.93 mg/L range during April, and (b) the 0.00-0.15 mg/L range during July-September. The trend noted in bias--increasing positive bias with increasing phosphorus concentration--probably reflects reagent limitation remaining in method I-2600/I-4600 even when: (a) colored and/or sediment laden samples were diluted two to fivefold prior to initial determination, and (b) off scale samples were diluted tenfold prior to redetermination. As noted, the trend was observed for filtered, as well as unfiltered samples. For the unfiltered samples, the major cause of the suspected reagent limitation can be rationalized as resulting from incomplete digestion of phosphorus compounds associated with progressively increasing concentrations of organic and mineral particles in suspended sediment. For filtered samples, perhaps the trend results from phosphorus associated with progressively increasing amounts of colloidal-size organic and mineral particles in the filtrates. We considered the possibility that some of the bias in both unfiltered and filtered samples might result from the somewhat different techniques used for colorimetrically measuring the orthophosphate (see page 1 for descriptions of the measurement step). The possibility was rejected, because instrument calibrations would have compensated for such differences in analytical results. Ground-water samples comprised roughly 5 and 10 percent of the July-September and April samples, respectively. The results of statistical testing for ground-water samples (Table 5) showed that: 1. In all comparisons, the estimated median bias is below 0.01 mg/L--the estimated method detection limit for I-2610/I-4610). 2. In general, the estimated median bias values are lower than observed for surface-water samples. 3. For certain comparisons, the bias is negative, rather than positive. 4. Statistically significant differences between the two methods are observed in only two of the seven cases. 5. No difference exists in bias for the filtered as compared to unfiltered samples. Although supportive data are lacking, Charles Patton believes that the small differences in total phosphorus concentrations between the two methods results from low concentrations of suspended sediment and/or organic carbon in the samples. For such samples, method I-2600/I-4600 would have little or no reagent limitation and, hence, would produce concentrations very similar to those from method I-2610/I-4610. IMPLICATIONS FOR DATA INTERPRETATION To summarize, two recent changes at the NWQL have affected analytical results for phosphorus. The first was a SOP change within method I-2600/I-4600 on May 1, 1990, that defined a set dilution procedure designed to ameliorate reagent limitation problems in samples having high concentrations of phosphorus, suspended sediment, and/or organic carbon. The second is the replacement of method I-2600/I-4600 with method I-2610/I-4610 on October 1, 1991. Before May 1, 1990, no SOP existed within method I-2600/I-4600 to dilute: (a) colored and/or sediment laden samples, or (b) off scale total phosphorus (> 1.0 mg/L) samples. For the former, there was no SOP to determine: (a) which samples arriving at the NWQL for total phosphorus analysis were colored or sediment laden enough to warrant dilution, or (b) volumetrically, how much to dilute the selected samples. Likewise, no SOP existed for the dilution of off scale dissolved phosphorus samples. Instead, dilution procedures were left to each analyst's discretion and, during the period 1980- 1990, over 25 analysts worked on the phosphorus line. In addition, NWQL had no system for recording whether individual samples were diluted and, if so, the selected dilution factor. Moreover, the data produced by the NWQL in 1990 to compare phosphorus measurements in high sediment samples by method I-2600/I-4600--with and without the dilution SOP--are insufficient in number to support statistical analysis. As noted, Tables 4 and 5 present the statistical results for comparing phosphorus measurements by method I-2610/I-4610 and method I-2600/I-4600 (with the dilution SOP). Additional data on the samples--such as concentrations of suspended sediment and organic carbon--are not available to enable adjustments for the detected biases. Tables 1-5 provide only a fragmentary picture of the bias in historic USGS phosphorus data. Tables 1-3 provide some insight to bias in USGS data produced prior to May 1, 1990, but relative to IEPA method 365.1, not to the subsequent USGS methods. Tables 4 and 5 show the estimated bias in the USGS data between the periods May 1, 1990 through September 30, 1991 and post October 1, 1991. For total phosphorus in surface waters, the combined results show: (a) a negative bias in USGS data produced prior to May 1, 1990, relative to IEPA data, and (b) a negative bias in USGS data produced during May 1, 1990 through September 31, 1991 relative to USGS data produced since using method I-2610/I-4610. Please note that the Illinois results provide only limited insight about the possible biases in USGS data produced prior to May 1, 1990. How well these results describe bias in samples from other U.S. rivers is uncertain due to the physical and chemical differences that may exist--such as the concentrations of sediment-bound phosphorus and organic carbon, and the nature and concentration of the suspended sediment. For dissolved (filtered) phosphorus in surface waters, Table 3 indicates that method I-2600/I-4600 (without the dilution SOP) did not produce a statistically significant bias relative to IEPA method 365.1. In contrast, the USGS methods comparison study on samples from across the country (Table 4) indicate that method I-2600/I-4600 (with the dilution SOP) caused statistically significant negative biases in dissolved phosphorus relative to method I-2610/I-4610. The bias for all filtered samples is 0.01 mg/L, compared to a bias of 0.02 mg/L for all unfiltered samples. However, the bias for unfiltered samples increases progressively with phosphorus concentration, and the individual biases for the high concentration ranges are similar to those for the comparable concentration ranges for total phosphorus. We believe that given the negative observed biases in the methods comparison study (Table 4), an even larger negative bias probably exists in USGS dissolved phosphorus produced before May 1, 1990. Perhaps the lack of bias in the Illinois filtered sample data and the presence of bias in the study comparing the USGS methods result from the digestion step in method I-2610/I-4610 being more rigorous (resulting in a more complete conversion of organic phosphorus compounds in the sample to orthophosphate) than in IEPA method 365.1. The OWQ will investigate this possibility by statistically comparing split sample phosphorus data now being produced by the IEPA and the USGS (using method I-2610/I-4610). Results will be communicated in a future OWQ Tech Memo. For ground-water samples, no split sample data are available to examine the possibility of bias in phosphorus measurements made prior to May 1, 1990. The results in Table 5 indicate that little or no negative bias exists in phosphorus data from ground-water samples between the periods of May 1, 1990 through September 30, 1991, versus post October 1, 1991. However, the number of ground- water samples in the methods comparison study was fairly small. As a result of the described situation, projects should recognize the following limitations in using historical (prior to October 1, 1991) USGS phosphorus data: Total Phosphorus in Surface-Water Samples 1. Total phosphorus data produced prior to October 1, 1991, tend to be biased low. This is especially true for samples having high concentrations of particulate phosphorus, suspended sediment, and organic carbon. It is likely that the negative bias is larger in the pre-May 1, 1990 data than in the data produced between May 1, 1990 through September 30, 1991. There is no general way to make scientifically defensible corrections to the data for either period. 2. Because the noted bias increases in samples having high concentrations of particulate phosphorus, suspended sediment, and organic carbon, estimates of annual loads of total phosphorus based on data produced prior to May 1, 1990, are likely to have a sizable negative bias. Negative bias is also likely in loads estimated from total phosphorus data produced from May 1, 1990 through September 30, 1991. 3. Interpretations of total phosphorus data across the time boundaries of May 1, 1990 and October 1, 1991 should be avoided. In the case of trend testing for concentrations, or worse, for estimated total annual loads, artificial upward trends could result because of the negative bias in total phosphorus measurements prior to May 1, 1990, and through September 30, 1991. 4. During statistical analysis, individual projects may be able to adjust for the bias in total phosphorus data, provided suitable historic split-sample data are available. The scientific integrity of bias-adjusted statistical estimates depends on the availability of split-sample data that are representative of the geohydrologic conditions of project sites. Ideally, split- sample data should exist for: (a) the observed ranges of discharges and the concentrations of particulate phosphorus, suspended sediment, and organic carbon; (b) the time period of interest; and (c) the sites of interest. In situations where site-specific data are not available, but geohydrologic characteristics are similar to those of the Illinois case study, the Illinois data may be useful for making approximate adjustments of bias for the pre-May 1990 period. However, large errors may be associated with such bias adjustments. Under no condition should bias adjustments be made to the individual data values in computer storage or in hard copy. The Branch of Systems Analysis may be contacted with questions about the use of statistical methods for bias adjustment. Dissolved Phosphorus in Surface-Water Samples 5. The same cautions cited in items 1,2, and 3 for total phosphorus generally apply to dissolved phosphorus. 6. As for total phosphorus, corrections of the bias in dissolved phosphorus data can be made for individual projects that possess the proper split sample data for the sites and time period in question (see item 4 above). Total and Dissolved Phosphorus in Ground-Water Samples. 7. Despite uncertainties, we believe that historic USGS phosphorus data for ground water have little or no negative bias. This belief is based on : (a) the limited results in Table 5, and (b) the fact that most ground-water samples have low concentrations of suspended sediment and relatively low (compared to surface waters) concentrations of organic carbon. QUALITY ASSURANCE/QUALITY CONTROL PROCESS This memorandum underlines the need for: (a) rigorous documen- tation and review of new methods and SOPs at the NWQL, and (b) use of field QC samples in all water-quality programs and projects. In 1990, the Division instituted a revised review process for implementing new analytical methods at the NWQL which requires: 1. Documentation of new methods in an Open-File Report, 2. Review of the manuscript by at least one analyst from another agency, and 3. Sign off of new methods by the Chief, Office of Water Quality. In addition, the Inorganics Program at the NWQL has established an approach for implementation of new SOPs that includes: documentation of plans; request for supervisor approval; review by supervisors; decision whether the change should be made or not; decision whether the change is major enough to warrant comparison testing; and record keeping on the chronology of changes. The Organics Program will soon implement a similar approach. Regarding field QC, the present system of Standard Reference Water Samples and Blind Samples did not detect the type of methodolog- ical error (reagent limitation) which produced low total phosphorus results, most notable on samples having high suspended sediment concentrations. Instead, as noted, the low total phosphorus results and, subsequently, the error in method, were detected by statistical interpretation of data from split samples generated when aliquots of the same field samples were analyzed by another laboratory. Available Standard Reference Water Samples for nutrients do not contain solids because reference samples must be stable for long time periods, and inclusion of solids tends to destabilize the nutrients, thereby significantly reducing shelf life. The Office of Water Quality is preparing QC guidelines to define the minimal percentage of various types of QC samples (for example, replicate samples, split samples, blanks, spikes, etc.) to include in projects for different combinations of objectives, sample matrices (water, sediment, tissue), and classes of chemical constituents. The intent is for projects to produce enough QC data to determine the quality of their environmental data and to identify and correct problems quickly. In addition, the Branch of Quality Assurance (BQA) is developing the capability to synthesize and interpret QC data from all Division programs to identify and resolve problems that may exist, but can not be detected by individual programs or projects reviewing their own data. Details on: (a) the QC guidelines, and (b) Division-wide analysis of QC data will be provided in future Tech Memos from the OWQ and BQA. REFERENCES Bowman, G.T., and Delfino, J.J., 1982, Determination of total Kjeldahl nitrogen and total phosphorus in surface waters and wastewaters: Journal of the Water Polution Control Federation, v. 54, p. 1324-1330. Jirka, A.M., Carter, M.J., May, Dorothy, and Fuller, F.D., 1976, Ultramicro semiautomated method for simultaneous determi- nation of total phosphorus and total Kjeldahl nitrogen in wastewaters: Environmental Science and Technology, v. 10, n. 10, p. 1038-1044. Murphy, J., and Riley, J.T., 1962, A modified single solution method for the determination of phosphate in natural waters: Analytica Chimica Acta, v. 27, p. 31-36. David A. Rickert Chief, Office of Water Quality Attachments Key Words: Methods, National Water Quality Laboratory, Phosphorus This memorandum does not supersede any previous Office of Water Quality Technical Memorandum. Distribution: A, B, S, FO, PO Table 1. Wilcoxon Signed Rank test results for seven stream sites comparing differences in total phosphorus (P) measurements by Illinois EPA (IEPA) method 365.1 and USGS method I-2600/I-4600 (without the dilution SOP) ___________________________________________________________________________ Median USGS Number total P IEPA suspended IEPA of (IEPA-USGS) (mg/L) (mg/L) (mg/L) samples (mg/L) p ___________________________________________________________________________ 0.05 - 2.00 147 0.010 <0.001* 0.05 - 0.20 42 0.010 0.255 0.20 - 0.30 30 0.010 0.021* 0.30 - 0.40 34 0.030 0.044* 0.40 - 2.00 42 0.020 0.015* 10 - 4,000 126 0.010 <0.001* 10 - 25 28 0.010 0.577 25 - 50 27 0.005 0.476 50 - 75 24 0.010 0.251 75 - 100 9 0.038 0.028* 100 - 4,000 36 0.061 0.004* 3 - 25 77 0.010 0.006* 3 - 7 38 0.010 0.171 7 - 25 39 0.020 0.015* 10 - 25 15 0.090 0.054 0.20 - 2.00 100 - 4,000 29 0.090 0.003* 0.30 - 2.00 100 - 4,000 20 0.180 0.005* ___________________________________________________________________________ * Significant for alpha=0.05 Table 2. Station Wilcoxon Signed Rank test results for seven stream sites comparing differences in total phosphorus measurements by IEPA method 365.1 and USGS method I-2600/I-4600 (without the dilution SOP) ___________________________________________________________________________ Median Median Median USGS Median total P Number IEPA suspended IEPA difference Station Name of total P sediment TOC DP/TP* (IEPA-USGS) (USGS ID #) samples (mg/L) (mg/L) (mg/L) % (mg/L) p ___________________________________________________________________________ Little Wabash R (3381495) 21 0.27 74 12 40 0.020 0.019** Rock River (5446500) 14 0.25 67 - 41 0.010 0.028** Spoon River (5570000) 12 0.14 86 - 43 0.017 0.038** Embarras River (3345500) 17 0.17 138 5 50 0.020 0.103 Big Muddy River (5599500) 21 0.25 68 - 39 0.010 0.322 Illinois River (5543500) 43 0.44 26 7 75 0.000 0.285 Sangamon River (5583000) 19 0.31 43 - 74 0.000 0.776 ___________________________________________________________________________ * Ratio of IEPA dissolved phosphorus to total phosphorus ** Significant for alpha=0.05 Table 3. Wilcoxon Signed Rank test results for seven stream sites comparing differences in dissolved phosphorus measurements by IEPA method 365.1 and USGS method I-2600/I-4600 (without the dilution SOP) _______________________________________________________ Median dissolved P difference IEPA dissolved P Number (IEPA-USGS) (mg/L) of samples (mg/L) p _______________________________________________________ 0.01 - 1.30 147 0.000 0.532 0.01 - 0.10 46 -0.005 0.157 0.10 - 0.20 38 0.005 0.075 0.20 - 0.30 25 0.000 0.093 0.30 - 1.30 26 -0.015 0.253 _______________________________________________________ Table 4. Wilcoxon signed rank results for surface-water samples comparing differences in phosphorus measurements by USGS methods I-2610/I-4610 and I-2600/I-4600 (with the dilution SOP) UNFILTERED ___________________________________________________________ Median total P difference I-2600/I-4600 Number (I-2610/I-4610- total P of I-2600/I-4600) (mg/L) samples (mg/L) p ___________________________________________________________ April samples 0.00 - 3.00 319 0.020 0.000* 0.00 - 0.15 209 0.013 0.000* 0.15 - 0.30 64 0.033 0.000* 0.30 - 0.50 24 0.044 0.006* 0.50 - 3.00 25 0.056 0.011* July-September Samples 0.00 - 4.50 616 0.020 0.000* 0.00 - 0.15 358 0.003 0.000* 0.15 - 0.30 108 0.031 0.000* 0.30 - 0.50 63 0.048 0.000* 0.50 - 0.70 21 0.064 0.000* 0.70 - 1.00 27 0.080 0.000* 1.00 - 2.00 27 0.151 0.000* 2.00 - 4.50 18 0.301 0.000* ___________________________________________________________ FILTERED ___________________________________________________________ Median dissolved P difference I-2600/I-4600 Number (I-2610/I-4610- dissolved P of I-2600/I-4600) (mg/L) samples (mg/L) p April samples 0.00 - 0.93 46 0.012 0.001* 0.00 - 0.21 40 0.011 0.002* 0.22 - 0.93 6 0.039 0.529 July-September Samples 0.00 - 4.00 506 0.006 0.000* 0.00 - 0.15 387 0.001 0.146 0.15 - 0.30 48 0.022 0.000* 0.30 - 0.50 21 0.043 0.000* 0.50 - 0.70 12 0.067 0.038* 0.70 - 1.00 12 0.090 0.038* 1.00 - 2.00 16 0.150 0.001* 2.00 - 4.00 16 0.208 0.001* ____________________________________________________ * Significant for alpha = 0.05. Table 5. Wilcoxon signed rank test results for ground-water samples comparing differences in phosphorus measurements by USGS methods I-2610/I-4610 and I-2600/I-4600 (with the dilution SOP) _____________________________________________________________ UNFILTERED _____________________________________________________________ Median total P difference I-2600/I-4600 Number (I-2610/I-4610- total P of I-2600/I-4600) (mg/L) samples (mg/L) p April Samples 0.00 - 3.25 20 0.003 0.184 0.00 - 0.20 19 0.004 0.062 July-September Samples 0.00 - 1.65 43 -0.003 0.169 0.00 - 0.23 40 -0.005 0.017* ____________________________________________________________ FILTERED ____________________________________________________________ Median dissolved P difference I-2600/I-4600 Number (I-2610/I-4610- dissolved P of I-2600/I-4600) (mg/L) samples (mg/L) p April Samples 0.00 - 0.70 32 0.008 0.030* 0.00 - 0.21 30 0.008 0.060 July-September Samples 0.00 - 0.34 18 -0.001 0.918 ____________________________________________________________ * Significant for alpha = 0.05