Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

A NUMERICAL MODEL TO ESTIMATE THE SEDIMENT OXYGEN DEMAND OF THE PASIG RIVER Lawrence P. BELO*; Pag-asa D. GASPILLO Chemical Engineering Department, De La Salle University 2401 Taft Avenue, Malate, Manila Fax: +63-2-536-0260 E-mail: [email protected]; [email protected] Masaaki SUZUKI; Junjiro KAWASAKI Chemical Engineering Department, Tokyo Institute of Technology Tokyo, Japan

Abstract: Sediment oxygen demand (SOD) is the rate at which dissolved oxygen is removed from the water column in surface waters mainly due to the respiration of benthic organisms and decomposition of organic matter in the riverbed or bottom sediments. Studies have shown that SOD can contribute from 30 to 90% of the total oxygen uptake especially in shallow and slowmoving waters. In a slow moving river with highly organic sediment like the Pasig River, SOD can be a major cause for the constantly low dissolved oxygen (DO) concentrations in the water column particularly in the summer months. Data collected from selected stations of the Pasig River were used to develop the empirical model to predict sediment oxygen demand levels. Several factors affect the variability in measuring SOD. Regression analysis was done in order to isolate the significant parameters directly affecting SOD. Correlation, regression analysis and ANOVA were then performed and the principal variables (S = TSS, %C = organic content and Q = flowrate) left lead to an empirical SOD equation: SODT  0.035(S )  27.9973(%C )  0.050348(Q)  2.55263 The SOD empirical equation was found useful in estimating the SOD of the stations although errors measured from as low as 0% to as high as 177% in measured SOD. The high variability between the predicted and the measured SOD were attributed other variables such as algal and biological population, BOD and COD, salinity and seasonal variability. Key Words: Sediment Oxygen Demand; Benthic Oxygen Demand; SOD Model; Pasig River; DO Uptake

1. INTRODUCTION The dissolved oxygen status of a water body is one of the most important parameters in assessing water quality. The demand for dissolved oxygen may be divided into two separate but highly interactive fractions: Sediment Oxygen Demand (SOD) and Water-Column Oxygen Demand (WOD) (Chau, 2002; Chen, 2000; Higashino, 2004; Steeby, 2004). Sediment Oxygen Demand is defined as the rate at which oxygen is removed from the water column in surface waters mainly

1

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

due to the respiration of benthic organisms and decomposition of organic matter in the bottom sediments taking place near the sediment-water interface. It includes both the aerobic respiration of the bottom organic substances, which is termed as the Biological SOD (B-SOD), and the chemical oxidation of reduced chemical species in the sediment, known as the Chemical SOD (CSOD) as illustrated in Figure 1 (Chau, 2002; Hantush, 2006; Hu, 2001). SOD greatly affects the balance of oxygen in water bodies and could account for majority of the overall oxygen uptake of the water body, and may reach from 30% to 90% of the total respiration especially in shallower estuaries (Belo et al., 2008; Butts and Evans, 1978; Chau, 2002; Parr, 2004).

Figure 1. The Overall Biological, Chemical and Geophysical Pond Reactions showing Oxygen Demand (-O2) and Oxygen Supply (+O2). In shallow water systems like the Pasig River, the water column and sediments interact continually. The Pasig River and its tributaries are of considerable interest since lowering water levels during the dry months can potentially enhance the net effect of SOD on water column dissolved oxygen by increasing the sediment bed area to lake volume ratio. Low tidal flushing, long residence time and vertical stratification in the Pasig River during these months can promote hypoxic conditions that may persist for long periods of time over large areas. In this regard, it is imperative that the status of bottom sediments be also monitored. However, despite its potential influence on the river’s dissolved oxygen budget, little work has been done in the Philippines in order to measure the SOD of Pasig River. The SOD data obtained in this study can be one of many input coefficients needed for water quality models which simulate the effect of an organic waste load on the river’s dissolved oxygen concentrations. In many cases, water quality modelers use a value of SOD = 0 because of lack on information on the potential importance of the SOD rate or due to limitation in the time to collect (Hatcher, 1987). In the case of the Pasig River, Qian et al. (2000) used a constant SOD of 1.6 g/m2d for their DO model to represent the whole stretch of the river. Hatcher (1987) stressed out that it is important for water quality modelers to have an accurate value for SOD to be used in 2

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

water quality models especially when dealing with water bodies with persistent bottom water hypoxia. Furthermore, SOD data obtained can be used in the implementation of subsequent water / wastewater rules and regulations. Moreover, considering the impact that the sediments contribute to the overall oxygen uptake in the river, SOD can be an additional criterion for surface water characteristics in the Philippine Clean Water Act and other water regulations like Department of Environment and Natural Resources (DENR) Administrative Order 34 (DAO 34).

2. METHODOLOGY 2.1 Study Site The Pasig River is one of the major rivers in the Philippines and connects two large important water bodies in Metro Manila namely, Laguna de Bay (a freshwater ecosystem) and Manila Bay (a saltwater system) (Figure 2). Since Metro Manila lies in the subtropical monsoon, rainy and dry seasons are clearly separated; December up to the following May is the local dry season where the river’s flow is low and water quality worsens accordingly (Qian et al., 2000).

Figure 2. The Pasig River map with sampling stations indicated. According to DAO 34 fishery waters, the Pasig River is classified as Class “C” wherein the BOD concentration should not exceed 7 mg/L as a yearly average and 10 mg/L as a maximum value and that DO should be 5 mg/L. (Figure 3) Regular monitoring by the government shows that average BOD is 10-30 mg/L; however during dry season when the depth of the river is extremely low, BOD can reach up to 100 mg/L. Even not in the extreme case, DO approaches to zero in almost all reaches of Pasig River, and the river becomes biologically unfit in dry season. Hence, these were the considerations for the three sampling river stations chosen for the study; Marikina (upstream), Jones (downstream) and Sanchez (middle stream) stations.

3

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

40 40 35 35 (mg/L) BOD(mg/L) BOD

30 30 25 25 20 20 15 15 10 10 55 00 1999 1999

Marikina Marikina

2001 2001

Bambang Bambang

2003 2003 Year Year

Lambingan Lambingan

2005 2005

Sanchez Sanchez

2007 2007as asofof2nd 2nd Quarter Quarter Jones Jones

DENR DENR

Figure 3. Annual Average Dissolved Oxygen and Biochemical Oxygen Demand Concentrations of the Pasig River stations from 1999 to 2007. 2.2 Sediment Oxygen Demand Measurement In the determination of the SOD rates used in the regression equation/model, laboratory runs were performed using bench-scale benthic respirometers (Figure 4) in monitoring the changes in the concentrations of dissolved oxygen of the given volume of water recirculated above a known area of sediment for a period of 2.5 hours.

Figure 4. Schematic Diagram of the benthic respirometer / SOD chamber The rates of oxygen loss were expressed as Sediment Oxygen Demand (SOD) and expressed in terms of grams of O2 per unit area of sediment per unit time (g O2/m2day) using the general SOD equation by the U.S. Environmental Protection Agency (US-EPA).

SODT  1.44

V b A

(Eq. 1)

where: SODT = V = A =

SOD (g O2/m2d) measured at ambient water temperature, T in 0C volume of isolated water circulated in the chamber (liters) area of the bottom sediments (m2)

4

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

Δb

=

1.44

=

difference in slopes of DO usage curves from blank chamber and the test chamber (O2/L-min) units conversion constant

Ambient SODT values were corrected to a reference temperature using the modified van’t Hoff form of the Arrhenius equation.

SODTr 

SODT 1.065T Tr

(Eq. 2)

where: SODTr = 1.065 =

SOD corrected to the reference temperature (200C and 250C) dimensionless model coefficient for temperature correction (Berthelson et al., 1996)

Temperature correction is necessary since this removed the effect of temperature on the measured SOD rates which is essential for comparison purposes. In addition, physical and chemical parameters of the water and sediment samples were also measured and tested against SOD for correlation. 2.3 SOD Regression / Numerical Model Hatcher (1987) stated that regression equations for estimating SOD rates can be developed by statistically correlating measured SOD rates with measured site conditions such as sediment depth, sediment chemical or physical properties, water quality and benthic quality parameters to determine if a relationship exists between the easily measurable parameters and SOD. With this, an attempt to come up with an SOD regression equation was done in the study. Stepwise multiple regression techniques were employed in an attempt to isolate the principal variables and to formulate them into a usable empirical predictive equation. The independent variables used in the regression analyses were: 1) Temperature, 0C; 2) Water Flow, ml/s; 3) Sediment depth in chamber, cm; 4) Organic Content, %w C; 5) Total Suspended Solids, mg/l; and 6) Volume of isolated water, L. SOD (g/m2day) values used in the statistical analysis were obtained at ambient temperature.

3. RESULTS AND DISCUSSIONS Simple correlation coefficient matrix between all the variables generated during the statistical analyses of the data is presented in Table 1. Focusing on the first column in the matrix, it can be noticed that with a large sample size of 76 only three out of the six measured variables were found to have high positive correlation to SOD, TSS (mg/l) being the highest, followed by organic content (% C) and flowrate (ml/s). Complete regression statistics are presented in Table 2 and Table 3. Table 1. Simple correlation matrix of measured variables against SOD

5

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

SODT Temp TSS %C flow sed depth volume

SODT

Temp

TSS

%C

flow

1 -0.00106 0.486339 0.311971 0.245118 -0.10444 0.104426

1 -0.39242 -0.57238 0.067468 -0.14601 0.145991

1 0.912863 -0.00664 -0.00734 0.007319

1 0.00042 -0.01266 0.012642

1 -0.03153 0.031501

sed depth

volume

1 -1

1

Although temperature was said to highly influence SOD, temperature differences of 1 to 2 degrees Celsius within a season would not have any significant effect on measured SOD rates (Butts, 1974). Total suspended solids and percent carbon are in strong agreement with SOD since both are indicators of the quality of water and sediments in the pond. The flow also exhibited a significant correlation with measured SOD since flow is inversely proportional to the thickness of the boundary layer that directly affects transfer of oxygen into the sediments. In the design of equipment, only two depths (2.5 and 5 cm) were investigated which led to only two variations in volume, 26.05 and 23.72 L for the low and high depths, respectively; both parameters did not yield significant correlations with measured SOD. Table 2. Regression Analysis and ANOVA for SOD and input parameters (Temperature, TSS, Organic Content, Flowrate, Sediment Depth and Volume) SUMMARY Regression Statistics Multiple R 0.644508154 R Square 0.415390761 Adjusted R Square 0.364555175 Standard Error 1.463704106 Observations 76 ANOVA df Regression 6 Residual 69 Total 75

SS 105.0380936 147.8276499 252.8657435

MS 17.50634893 2.142429709

F 8.171259

Significance F 1.1E-06

Coefficients -4986.360848 -0.03436403 0.035598122 -29.05420265 0.050670039 163.4971487 175.6648955

Standard Error 32694.68557 0.222110038 0.007063448 9.763223752 0.018776587 1073.26245 1152.012876

t Stat -0.152512886 -0.154716241 5.039765203 -2.975882085 2.698575574 0.152336596 0.152485184

P-value 0.879228 0.877497 3.59E-06 0.004025 0.008748 0.879366 0.879249

Lower 95% -70210.5 -0.47746 0.021507 -48.5313 0.013212 -1977.6 -2122.54

Intercept Temp TSS %wt C Flowrate Depth Volume

6

Upper 95% 60237.75 0.408733 0.049689 -9.5771 0.088128 2304.597 2473.868

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

Multiple regression analysis was applied to the data obtained. After determining the significant input variables from Table 2, another set of regression as presented in Table 3 was performed on the remaining data to obtain the predictive SOD equation. Table 3. Regression analysis and ANOVA for the TSS, flowrate and % Organic Content versus SOD SUMMARY OUTPUT Regression Statistics Multiple R 0.636643 R Square 0.405314 Adjusted R Square 0.380536 Standard Error 1.445182 Observations 76 ANOVA df Regression Residual Total

Intercept TSS %C FLOW

3 72 75 Coefficients -2.55263 0.035003 -27.9973 0.050348

SS 102.4901 150.3756 252.8657

MS 34.16337 2.088551

Standard Error 0.754515 0.006388 7.774607 0.018049

t Stat -3.38314 5.479261 -3.60113 2.789478

F 16.35745

P-value 0.001162 5.96E-07 0.000579 0.006751

Significance F 3.32E-08

Lower 95% -4.05673 0.022268 -43.4957 0.014368

Upper 95% -1.04853 0.047737 -12.4989 0.086329

From the correlation, multiple regression analysis and ANOVA (with 95% confidence) done on the data, only the significant variables (TSS, %C, flow) were retained in the predictive SOD equation for the Pasig River. The formula for SOD at ambient Temperature, T, is as follows:

SODT  0.035(S )  27.9973(%C )  0.050348(Q)  2.55263

(Eq. 3)

where: SODT S %C Q

= SOD measured at ambient temperature, T (0C) = Solids as Total Suspended Solids concentration (mg/L) = Organic content as percentage carbon (%C) = flowrate (ml/s)

The multiple R value of 0.64 and a standard error of 1.44 were obtained in the predicted SOD equation. When only the significant variables were retained and analyzed in a new equation, the correlation increased thus minimizing the standard error. An investigation on possible interactions between parameters were also performed in the regression model generation; however, in doing so, the only significant parameter left is the flow leaving all other parameters insignificant. Transforming the carbon content and TSS to log base 10 increased the correlation

7

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

coefficient relating each parameter to SOD, which were also observed by Butts (1974) and Butts and Evans (1978) in their studies. However when multiple regression was performed on the transformed form, the multiple R values obtained were either lower or just equal to the R value in the original linear form (Eq. 3). Hence, for ease in manipulation, the linear form (Eq. 3) was chosen by the researcher as the predict SOD equation for the Pasig River.

4. CONCLUSIONS The predictive equation was found helpful in estimating the SOD of the Pasig River. The three variables (S, %C and Q) in Equation 3 accounted only for 41% of the overall variability and the other three (T, sed depth and volume) accounted for only 1%. Consequently, the remaining 58% of the measured variability in predicted SOD from the observed SOD could be attributed to other unknown factors which are probably related to algal and biological populations, BOD and COD, salinity and seasonal variability; nevertheless, determination of the influence of these factors is beyond the scope of this study. Estimated SOD using the numerical model generated errors from as low as 0% to as high as 177% in measured SOD. It has been stated, however, that no particular regression equation was found to be generally applicable for estimating SOD rates at all sites since regression equations developed for one water body may not apply to other dissimilar water bodies with different hydrologic conditions (Hatcher, 1987). Nevertheless, regression equations developed for a category of water bodies can be useful for estimating SOD when the modeler has little or no SOD field data but has the parameters in the regression equation.

REFERENCES Belo, L.P., P.D. Gaspillo, J.M. Lim, J. Kawasaki and M. Suzuki (2008) Master’s Thesis: Measurement of the Sediment Oxygen Demand in Selected Stations of the Pasig River Using a Bench-Scale Benthic Respirometer. Butts, Thomas A. (1974) Measurement of Sediment Oxygen Demand Characteristics of the Upper Illinois Streams. Illinois State Water Survey, Urbana, Report of Investigation 76. Butts, T.A and Evans, R.L. (1978) Sediment Oxygen Demand Studies of Selected Northeastern Illinois Streams. Illinois State Water Survey, Urbana, Circular 129. Chau, K.W. (2002) Field Measurements of SOD and sediment nutrient fluxes in a land locked embayment in Hong Kong. Advances in Environmental Research vol. 6, p. 135-142. Chen, G.H. et al. (2000) Oxygen deficit determinations for a major river in eastern Hong Kong, China. Chemosphere, Vol. 41, pp 7-13. Hantush, Mohamed (2007) Modeling nitrogen–carbon cycling and oxygen consumption in bottom sediments. Advances in Water Resources, Vol. 30, p.59-79. 8

Presented at the 1st Regional Conference in Chemical Engineering Century Park Hotel, Manila, Philippines January 22 – 23, 2009

Hatcher, Kathryn J. (1987) Selecting an Appropriate Method for Estimating the Sediment Oxygen Demand Rate: Chemical and Biological Characterization of Municipal Sludges, Sediments, Dredge Spoils and Drilling Muds, p. 438-449. Higashino, Makoto et al. (2004) Unsteady diffusional mass transfer at the sediment-water interface: Theory and significance for SOD measurement. Water Research vol. 38, p.1-12. Hu, W.F. et al. (2001) Nutrient release and sediment oxygen demand in eutrophic land-locked embayment in Hong Kong. Environment International, Vol. 26, p.369-375. Parr, Lynn and Christopher Mason (2004) Causes of low oxygen in a Lowland, Regulated Eutrophic River in Eastern England. Science of the Total Environment, Vol. 321. p.273-286. Qian, X., Capistrano, E., Lee, W. and Ishikawa, T. (2000) Flow Structure and Water Quality of Pasig River in Metro Manila. Korea Water Resources Association (KWRA), International Conference on Hydro-Science and Engineering in 2000. ICHE-2000. Steeby, James A. et al. (2004) Modeling Industry-wide sediment oxygen demand and estimation of the contribution of sediment to total respiration in commercial catfish ponds. Aquaculture Engineering 31, p.247-262. Walpole, R. et al. (2003) Probability and Statistics for Engineers and Scientists, Seventh Edition. Prentice Hall, New Jersey.

9