Assessing, Sampling and Analysis Techniques

Do water samples collected routinely for monitoring programmes accurately reflect river phosphorus (P) and suspended solids (SS) concentrations? This paper examines several stages of standard sampling, preservation and analysis techniques (SSPAT) for water samples from two lowland UK rivers. Although universal analytical procedures are necessary for data comparison, this paper indicates that adopting a SSPAT approach alone may jeopardise sample representativeness. Therefore, preliminary surveys to assess whether SSPAT protocol is sufficient to quantify P and SS loads are highly recommended.

Introduction

To assess nutrient and sediment mobility within rivers, P fractions and SS concentrations are normally determined on samples collected from the water column 1 . Recommendations for sampling frequency 1 and shortfalls in the laboratory analysis of P 2 are well documented in the literature. However, there are several stages of SSPAT that need more consideration before initialising a P and SS monitoring programme. Loads are derived from the product of concentration and discharge over a specified time increment. However, whilst determining discharge does not normally present any practical problems obtaining representative concentration data is more problematic. Nutrient and sediment loads in river basins are under increased scrutiny in the EU community because the deadline for meeting Water Framework Directive objectives will be implemented in 2015.

A crucial question is: do water samples collected routinely in P and SS monitoring programmes accurately reflect actual concentrations in the river? If not, what degree of uncertainty does this generate in loads estimated from these programmes? This paper addresses some of the limitations involved in adopting SSPAT protocol in terms of assessing P and SS mobility within two lowland UK catchments.

Sample sites

The Rivers Lambourn and Enborne are tributaries of the River Kennet, England (Figure 1). The Lambourn lies on Chalk and has a fast flowing, shallow channel with extensive water crowfoot beds (Ranunculus penicillatus var. calcareous). The regime of the Lambourn is subdued with hydrographs dominated by delayed through flow and maximum flow in March. The Enborne drains impermeable Tertiary sand, silt and clay deposits. The river has a slow flowing, deep channel with limited submergent macrophytes. The regime of the Enborne is flashy, with maximum monthly flows in December/January, coinciding with maximum precipitation, and hydrographs are dominated by quick flow response.

Two 500 metre reaches were selected as sampling sites for this study; Boxford on the Lambourn (national grid reference SU429719) and Brimpton on the Enborne (national grid reference SU568468) (Figure 1).

Spatial variation in P and SS

P and SS concentrations exhibit a wide degree of spatial variability along a river continuum at any given time. However, many studies neglect to investigate whether this variation affects data sampled from a single point and assume that they will be ‘averaged’ out over a long term monitoring programme. Kronvang (1996) observed that the accuracy obtained by point sampling compared to the rest of the cross section in the River Gjern (Denmark) was relatively high for SRP and PP (mean variation 1.1% and 4.9%, respectively). However, in larger catchments cross-sectionally integrated sampling is recommended 3 .

To assess spatial variation in the two river reaches 10 sampling points were selected within the upstream and downstream limits of each reach of the Rivers Lambourn and Enborne (Figure 2, positions 1 to 10). A manual grab sample was collected simultaneously from each of these positions on three occasions (Spring 1998, Summer 1999, Winter 1999) to reflect seasonal differences in the flow domain of each river and the mobilisation and transport of P and SS in each catchment. These samples were analysed for SS and full P fractionation.

The variations in SS and P fraction concentrations at the ten sampling points in the River Lambourn and Enborne study reaches are shown in Figure 3. In the River Lambourn, SRP and Soluble Unreactive Phosphorus (SUP) had low spatial variation due to mixing by the turbulent flow. PP and SS had higher spatial variation because particulate matter was deposited in some areas of the channel (particularly around the plant beds) and re-suspended in others. Spatial variations for both the dissolved and particulate fractions of P were lower in the River Enborne. This reflected the more homogenous channel character (i.e. low velocity and sediment deposition variability).

The differences in SS and P fraction concentrations for the two River Lambourn cross-sections were analysed using the F-test (Table 1a). There were no significant differences in SRP, SUP and PP concentrations collected between positions 1 and 2-4 and none in SRP and SUP concentrations collected between positions 7 and 8-10. Bridges upstream from both the cross-sections and high flow velocities observed in the Lambourn would increase the turbulent flow patterns within the channel. The statistical analyses revealed that there was a significant difference between SS concentrations collected between positions 1 and 2-4 and both SS and PP concentrations collected between positions 7 and 8-10. Patches of streambed colonised by macrophyte beds and differences in water velocity and bed shear stress clearly affected particulate transport in the cross-section.

The differences in SS and P fraction concentrations for the two River Enborne cross-sections were also analysed using the F-test (Table 1b). The statistical analyses demonstrated that there were no significant differences in SRP and PP concentrations collected between positions 1 and 2-4 or in SRP, SUP and PP concentrations collected between positions 6 and 7-9. This is probably due to the uniformity of the channel cross-sections with regard to flow velocity and channel morphology.

The statistical analyses also revealed that there were significant differences in SS and SUP concentrations collected between positions 1 and 2-4 and in SS concentrations collected between positions 6 and 7-9. These spatial variations were probably due to incomplete mixing between the main river channel and a drainage ditch that flowed into the Enborne upstream from the cross-section. Spatial variation data has shown that soluble P fractions suffer least from in-channel differences. Data collected by a fixed-point auto sampler might, therefore, be regarded as being representative of these cross-sections and would be reliable for calculating the loads of SRP and SUP transported in the Rivers Lambourn and Enborne or other rivers of moderate size (cross-sectional area 10-15 m 2 ).

However, these manual samples were collected during periods of base flow in March and September. Non-synchronous changes in the behaviour of the cross-section with respect to water column chemistry may occur during storm flow events and the macrophyte growing season. Particulate matter concentrations appeared to vary greatly within the channel of both rivers so any estimate of SS and PP loads should be treated with a degree of caution.

However, the mean SS concentrations in the River Lambourn at position 1 and 2 – 4 were the same – it was the spatial variation at positions 2 – 4 that samples from position 1 were not representing. Data from a fixed-point may, therefore, provide a statistically robust estimate of loads of SS and PP passing the cross-section on an annual basis. However, it will not detect peaks and troughs in SS and PP at all the positions along the cross-section during storm events when the majority of fine sediment transport occurs 1 .

Automatic water sampling

There has also been little work which evaluates the difference in the P chemistry and SS concentrations of samples collected manually and automatically. Automatic samplers have the advantage that they can be deployed in the field for long periods of time and are thus more likely to capture infrequent high flow events that can transport disproportionately high P and SS loads.

Triplicate automatic and manual water samples collected simultaneously from the same position (positions 1 and 7, Figure 2) were analysed for SS and full P fraction concentrations to provide a comparison between the two modes of collection. Epic peristaltic pump auto samplers were used at all locations. In addition, double deionised ultra pure water was sampled on-site at each automatic station before the sampling programme started and again after two years. Again, these samples were analysed for full P fractionation to assess the degree of contamination in the auto samplers after long term deployment.

Table 2a shows that SS and PP concentrations were consistently lower in manual samples compared to automatic samples in both the Lambourn and Enborne (mean differences of 2 mg l -1 and 0.019 mg P l -1 , respectively). However, no statistical differences were found for SRP and SUP concentrations between manual and automatic sampling (mean differences of 0.004 mg P l -1 and 0.005 mg P l -1 , respectively). These values are below reported limits for analytical error of P analysis. No sample contamination was observed using auto samplers prior to the start of the sampling program (Table 2b). However, low concentrations of P (range 0.002 – 0.034 mg P l -1 ) were discovered when pumping double distilled water after the field program had finished. In these contaminated samples, P fractionation was dominated by PP suggesting that solids had accumulated in the sampling pipes or sample chambers during deployment.

The differences found in both SS and PP concentrations collected by manual and automatic samplers could be due to fouling of the auto sampler pipe and sample chamber by algae, insufficient lift of previous samples with high SS content or resuspension of fine particulate matter from the streambed/auto sampler pipe support due to the pre-purge cycle on auto sampler operating programmes. These differences have important implications for estimation of particulate loads in base flow dominated streams with low sediment transport rates such as the Lambourn.

The size of the error between the two values will magnify the error in load calculations based upon auto sampler data in the Lambourn. In streams with larger sediment transport rates such as the Enborne, particulate loads will again be overestimated but the magnitude of error will be smaller. On the other hand, manual samples may underestimate true SS and PP loads due to disruption of flow patterns in the channel caused by the greater cross-sectional area of the sample bottle (compared to an automatic sampling tube).

In addition, manual samples collected using the traditional ‘bucket over the-bridge’ approach will also underestimate SS and PP concentrations because they sample a disproportionate amount of surface water with lower associated sediment loading than at an average channel depth.

Collection of blank samples identified auto sampler contamination by in-situ algal matter production and inorganic particles not purged from the system. Based upon this sampling error, PP loads could be under or overestimated. In addition, PP in contaminated samples left in bottles for extended periods might start to solubilise, leading to an error in the soluble P fraction load. It is, therefore, recommended that used auto sampler tubes are substituted for clean ones on a fortnightly basis (especially during the summer period). The length of tubing should be kept to a minimum to avoid deposition of heavier particles within the tube due to insufficient pump lift.

Sample preservation – Storage stability tests

Significant alterations in the SRP concentration of stored samples have been noted in many studies 6,7 . A number of sample treatments have been used to try to reduce these changes, including filtration and refrigeration, freezing or the addition of biocidal preservatives such as mercuric chloride, chloroform and sulphuric acid, with variable success7. This research has highlighted the need for storage stability trials for every river prior to the initiation of a P sampling programme to decide which storage technique is most appropriate.

To investigate sample instability, triplicate samples collected from the Rivers Lambourn and Enborne were all divided into six sub-samples (18 samples of 100 ml each). 50 ml of one sub-sample was filtered (0.45 μm pre-washed cellulose nitrate filters) immediately on-site and refrigerated at 4°C in dark storage. 50 ml of the remaining five sub samples were laboratory filtered within two hours of collection. 50 ml of the five paired filtered and unfiltered sub-samples were then treated in one of the following ways:

• Untreated control
• Addition of 2 ml mercuric chloride (40 mg Hg2+l -1 )
• Addition of 2 ml sulphuric acid (5%)
• Frozen (4 sub-samples to ensure each sample only frozen/thawed once)
• Stored in a box outside the laboratory (to mimic the conditions during a typical six day auto sampler deployment)

All the sub-samples were refrigerated at 4°C in dark storage (with the exception of the frozen paired sub-samples). These sub-samples were analysed for full P fractionation on immediate return to the laboratory to set a baseline level against which to compare future analyses on the preserved samples. Subsequent analyses were performed after one, seven, fourteen and twenty-eight days

Initial (Day 0) mean SRP, SUP and PP concentrations were 0.098, 0.010, 0.097 mg P l -1 for the Lambourn, respectively and 0.243, 0.038 and 0.155 mg P l -1 for the Enborne, respectively. Figure 4 shows that samples collected from the River Lambourn (mean monthly SRP percentage change -50%) were more unstable than those from the River Enborne (mean monthly SRP percentage change -18%). Overall, SRP was the most unstable fraction, followed by SUP then PP.

The most dramatic fluctuations in P concentrations were observed in samples preserved by mercuric chloride and freezing techniques. Percentage change data suggested conversion of PP and SUP into the SRP fraction. Samples left in an auto sampler performed favourably compared to other storage techniques over the first day (-7 and -5% change in SRP concentrations in Lambourn and Enborne samples, respectively). However, large decreases in all P fraction concentrations were noted over the one month period (-70 and -17% change in SRP concentrations in Lambourn and Enborne samples, respectively). P concentrations were most stable following on-site filtration with -2 and – 3% change in SRP concentrations in Lambourn and Enborne samples, respectively, over the first day.

Laboratory filtered samples showed greater change in P concentrations over the first day (-5 and -6% in the Lambourn and Enborne, respectively). However, subsequent analysis after seven, fourteen and twenty-eight days showed little difference between field and laboratory filtered samples.

P concentrations changed in stored samples from both rivers due to a combination of effects noted by previous authors such as desorption of PP, adsorption of P onto container walls, cell rupture, co-precipitation of P with calcium carbonate and hydrolysis of organic P. It was clear from storage stability testing that Lambourn water samples were less stable than Enborne samples and that adsorption onto bottle walls rather than conversion to other fractions of P was the key control upon SRP concentrations.

This was because P concentrations were lower in the Lambourn and PP concentrations were more stable in the Enborne; possibly due to the nature of sediment-P interactions. The most effective method for preserving River Lambourn and Enborne water samples was on site filtration followed by refrigeration in the dark and analysis within 24 hours.

Size-fractionated transport of PP

The transport of P by colloidal matter <0.45 μm has been noted previously 8 . However, routine P monitoring programmes still use the 0.45 μm filter as the standard operational cut-off point between particulate and soluble P fractions.

To assess the role of different sediment size fractions in transporting P in river systems, weekly water sub-samples from the Rivers Lambourn and Enborne over a period of 1.5 years were sequentially filtered through four different pore size filter papers (1.00, 0.45, 0.20 and 0.10 μm). Four sub-samples (rather than one sample) were used to reduce contamination risk. Each sub-sample was then analysed for TP. Large differences were observed between TP concentrations in the four filtrates of River Enborne samples (long-term means of 0.354, 0.304, 0.239 and 0.223 mg P l -1 for the 1.00, 0.45 and 0.20 and 0.10 μm filters, respectively). In River Lambourn samples, there was a smaller decrease in the concentration of the TP fraction with filter pore size (long-term means of 0.176, 0.148 and 0.144 mg P l -1 for the 1.00, 0.45 and 0.20 μm filtrates, respectively). This pattern for samples from both rivers can be attributed to a decrease of particles contained within the sub-samples at successively lower pore sizes. However, the long-term mean for the finest, 0.10 μm Lambourn filtrate (0.147 mg P l -1 ) was higher than for the next coarsest filter.

Statistical analysis of TP concentrations in 1.00, 0.20 and 0.10 μm filtrates compared to that of the 0.45 μm filtrate from the Enborne and Lambourn highlighted the operational differences between dissolved and filtered P (Table 3). Significant differences between the 1.00 and 0.45 μm filtrates were observed indicating the importance of fine sands, silt, algae and flocculated colloidal matter size particles in determining TP concentrations. No statistical differences were noted between 0.45 and 0.20 μm filtrates suggesting that bacteria size particles were not a significant component of TP concentrations in the Lambourn or Enborne. However, significant differences were observed between the 0.45 and 0.10 μm filtrates reflecting the role that colloidal matter might play in P transport in both rivers.

Changes in TP in Enborne and Lambourn samples revealed seasonal variations in P concentrations between the 1.00 and 0.45 μm filtrates. In the Lambourn large differences from November 1998 to February 1999 and again from December 1999 to March 2000 were observed. In the Enborne large differences between the 1.00 and 0.45 μm filtrate in the periods July to August 1999 and again from October 1999 to January 2000 were observed. TP from the 0.10 and 0.20 Enborne and Lambourn filtrates did not deviate largely from 0.45 μm filtrates. However, a closer agreement in concentrations during low flow in the summer and autumn could be attributed to reduced transport of fine sediments <0.45 μm during this period.

Progressively lower P concentrations observed in the four sequential filtrates in the Lambourn and Enborne were similar to data found by Haygarth et al., 1997. Cell rupture or mechanical destabilisation of flocs might be responsible for consistently higher P concentrations in the 0.1 μm filtrate (compared to the 0.2 μm filtrate) in the Lambourn due to an increase in pump pressure necessary for filtration. The larger differences between P concentrations in the Enborne (compared to the Lambourn) indicated that the choice of filter pore size would be more critical for River Enborne samples. Large differences between P concentrations in the 1.00 and 0.45 μm filtrates on a seasonal scale may have been due to the transport of epiphytes and organic matter within the channel during the summer/autumn period. Following this, increasing flows during the winter would transport coarser sediment in suspension resulting in larger differences in PP between the two filtrates. The wider implication of this work is that it is entirely feasible for P to have potential for being physically mobile (by being attached to colloids, silts and algae), whilst remaining chemically less active than soluble components of the water column. In particular, the role of colloidal matter in transporting P during base flow periods is probably very significant in terms of P export and productivity in the River Enborne. This challenges the traditional premise that fluvial transport is dominated by dissolved inorganic P in the summer.

Conclusions

The following four important observations were made during this work:

• Spatial surveys of the river channels found statistically significant variations in SRP and PP not accounted for by point sampling (± 0.005 and 0.013 mg P l -1 , respectively)
• Comparisons between manually and automatically derived samples found no significant differences in SRP, but PP and SS concentrations were consistently lower in manual samples (on average 0.019 mg P l -1 and 3 mg l -1 , respectively). Contamination of auto samplers during a two year long-term deployment was quantified (accounting for, on average, increases of 0.009 mg SRP l -1 and 0.015 mg PP l -1 )
• Storage stability testing indicated that P concentrations in water samples were susceptible to changes following collection. Immediate on site filtration, refrigeration and subsequent analysis within 24 hours was the optimum of six preservation techniques tested (resulting, on average, in a decrease of 0.007 mg SRP l -1 )
• Sequential filtration revealed statistically significant differences in P concentrations between 0.45 μm and 0.10 μm sub-samples (mean decrease of 0.041 mg P l -1 ). This illustrated the important role of fine particulates in P transport

In the past, precautionary practice has been used to facilitate comparison of P data from different catchment programmes by setting up SSPAT protocol. However, good practice would also include site specific measurements geared to the SSPAT protocol but taking local river conditions into account 3 . These detailed measurements are essential to the accurate determination of P and SS loads. For regional, national and international scale monitoring, we need to learn from these findings to guide us to the best techniques available, whilst accepting that site specific validation testing is not possible in all situations.

References

• Kronvang B, Bruhn AJ (1996) Choice of sampling strategy and estimation method for calculating nitrogen and phosphorus transport in lowland streams. Hydrological Proc, 1483-1501. Kronvang B, Bruhn AJ (1996) Choice of sampling strategy and estimation method for calculating nitrogen and phosphorus transport in lowland streams. Hydrological Proc, 1483-1501.
• Neal C, Neal M, Wickham H (2000) Phosphate measurement in natural waters: two examples of analytical problems associated with silica interference using phosphomolybdic acid methodologies. Sci Total Environ, 511-522.
• Jarvie HP, Withers PJ (2002) Phosphorus sampling, storage and analysis. In: Kronvang B (Ed.), Diffuse phosphorus losses at catchment scale. Proc. WG3 COST action 832 workshop 26-28 October 2000, Silkeborg, Denmark.
• Manning (2005) Guide to vacuum vs. peristaltic fluid samplers. Available online at [www] www.manning-enviro.com/guide.php (accessed 22/02/2005).
• Evans JG, Wass PD, Hodgson P (1997) Integrated continuous water quality monitoring for the LOIS river programme. Sci Total Environ, 111-118.
• Bull KR, Lukhani KH, Rowland AP (1994) Effects of chemical preservative and temperature storage conditions on cations and anions in natural water. Chemistry and Ecology, 47-62.
• House WA, Warwick MS (1998) Intensive measurements of nutrient dynamics in the River Swale. Sci Total Environ, 111-137.
• He ZL, Wilson MJ, Campbell CD, Edwards AC, Chapman SJ (1995) Distribution of phosphorus in soil aggregate fractions and its significance with regard to phosphorus transport in agricultural runoff. Water, Air and Soil Pollution, 69-84.
• Haygarth PM, Warwick MS, House WA (1997) Size distribution of colloidal molybdate reactive phosphorus in river waters and soil solution. Water Res, 439-448.
• Stevens RJ, Smith RV (1978) A comparison of discrete and intensive sampling for measuring the loads of nitrogen and phosphorus in the River Main, County Antrim. Water Res, 823-830

Published: 10th Dec 2008 in AWE International