1897
© IWA Publishing 2014 Water Science & Technology

70.12

2014
Further contributions to nitrogen removal modelling in waste stabilization ponds R. K. X. Bastos, V. A. L. Cabral, E. N. Rios and M. P. M. Combatt
ABSTRACT A large database from an experimental maturation pond system in Brazil was used to verify the agreement of ﬁeld results with values predicted by some of the most widely accepted models to describe ammonium and total nitrogen (TN) removal in facultative and maturation ponds. The same database was used to derive a pHindependent linear model to predict ammonium removal in ponds, which was proved to be, essentially, a function of ammonium surface loading rate. In general, all these models made reasonable predictions of ammonium or TN removal but tended to overestimate low ammonium efﬂuent concentrations while underestimating higher values of ﬁeld data. Key words
 maturation ponds, modelling, nitrogen removal
R. K. X. Bastos (corresponding author) V. A. L. Cabral E. N. Rios Departamento de Engenharia Civil, Universidade Federal de Viçosa, ViçosaMG, 36570000, Brazil Email:
[email protected] M. P. M. Combatt Departamento de Tecnologia de Alimentos, Universidade Federal de Viçosa, ViçosaMG, 36570000, Brazil
INTRODUCTION Pano & Middlebrooks’ () models have been widely used for predicting ammonium concentration in waste stabilization ponds (WSP) efﬂuents, generally under the assumption that ammonia volatilization is the main mechanism for nitrogen removal. However, as pointed out by Camargo Valero & Mara (), the Pano and Middlebrooks models are actually simple ﬁrstorder equations for a completely mixed reactor, which reﬂex statistically signiﬁcant relationships with pH, water temperature and hydraulic loading rate, but do not provide any information on the nature of the nitrogen removal mechanisms. Similar observations apply to the models proposed by Reed () and Middlebrooks in 1985 (given in Middlebrooks et al. ()) for predicting total nitrogen (TN) removal, which were developed for plugﬂow and completely mixed conditions, respectively. Both these models are also dependent on pH, temperature and hydraulic retention time (HRT), in which it is also assumed that ammonia volatilization is the major pathway for nitrogen removal. Water temperature, pH, and hydraulic loading rate are indeed important factors for NH3 stripping (i.e. the unionized gaseous form or free ammonia), but are equally important factors for other biological processes in ponds such as an active primary productivity leading to algal nitrogen uptake or simultaneous nitriﬁcation–denitriﬁcation. In effect, as pH ﬂuctuates as a result of the algae–carbonate interactions in WSP, high pH values may simply reﬂect intense algal activity, hence algal uptake of the ammonium doi: 10.2166/wst.2014.365
ionic form (NHþ 4 ). Thus, and despite the fact that there have been more and more reports that ammonia volatilization makes only a small contribution to the overall performance of nitrogen removal in WSP (Camargo Valero & Mara ; Camargo Valero & Mara ; Assunção & von Sperling ), it seems that Pano and Middlebrooks’ models and slight variations of them still make good predictions of ammonium removal in WSP (Silva et al. ; Soares et al. ; Bastos et al. ). The aforementioned models were developed based on theoretical considerations for ammonia stripping in a completely mixed or a plugﬂow reactor under steadystate and continuous ﬂow conditions, along with other mechanisms such as ammonia removal through biological activity and ammonia releasing into the pond water column from anaerobic activity at the bottom of the pond. The ﬁrstorder kinetics models resulting from these theoretical considerations were then calibrated using data from speciﬁc conditions in terms of pond type, pond organic load and climate, and were further validated using data from sources other than those used in model development (Middlebrooks et al. ; Crites et al. ). Yet, as a WSP design tool, all these models suffer from the drawback of being dependent on pH, a variable not known at the designing stage. In this study a large database from an experimental maturation pond system in Brazil was used to further discuss the usefulness of these literature classical models for ammonium and TN removal in maturation ponds. In
1898
R. K. X. Bastos et al.

Nitrogen removal in waste stabilization ponds
addition, the same database was used in an attempt to derive a pHindependent model to predict ammonium removal.
METHODS A large database (as a whole, 699 and 296 results for NH3 and TN, respectively), obtained from monitoring an experimental sewage treatment plant (STP) in Brazil from October 2001 to December 2011, was used to verify the agreement of ﬁeld results with values predicted by the Pano and Middlebrooks model in terms of ammonium removal, and by both the Middlebrooks and Reed models for TN removal. The experimental STP, situated in Viçosa, Minas Gerais State, southeast Brazil (latitude: 20 450 14″S, longitude: 42 520 53″W, altitude: 650 m), comprised by upﬂow anaerobic sludge blanket (UASB) reactor, a submerged aerated bioﬁlter (BF) (ﬁeld scale) and a fourmaturation pond series (pilot scale). All ponds were 16.3 m2 and had length/width ratio equal to 2.0. Over the study period, the pond system was operated under varying conditions of climate, hydraulic surface loading rate (SLR) and HRT (Table 1). Actually, the pond system had three ponds up to 2003; hence the missing data for pond 4 in periods 1, 2 and 3 in Table 1 Also, due to ﬁeld constraints, the pond system monitoring was not continuous, thus the time lapses between some periods in Table 1. An attempt was made to ﬁt a variation from the original Pano and Middlebrooks model to the experimental data. In addition, attempts were made to derive a pHindependent model to predict ammonium removal. Aiming at having a large database, representative of varied operational conditions, data from each maturation pond were gathered in a single dataset, on which stepwise multiple linear regression analysis was applied using the software Minitab Pro 16.2. Firstly, regression analysis was performed between ammonium concentration in pond efﬂuent (Ce) and the following independent variables: (i) water quality variables – biochemical oxygen demand (BOD), pH, total suspended solids (TSS), chlorophyll a, and temperature (T); and (ii) operational variables – HRT, ammonium SLR, and pond depth (h). Following the evaluation of the statistical signiﬁcance (pvalue) and strength (coefﬁcient of determination R 2) of this ‘complete model’, the software’s ‘stepwise tool’ was used in order to obtain a parsimonious model entailing only the statistically signiﬁcant variables. In a further step, those variables which are not known at the design stage were withdrawn and another parsimonious model was obtained using stepwise multiple linear regression. The number of analysed W
W
Water Science & Technology

70.12

2014
data in each of these trials varied according to the results availability: 221 values of each variable involved in generating the ﬁrst ‘complete’ model, 476 and 572 values for the second and the ﬁnal model respectively. Finally, the statistical validity of the resulting model was assessed using graphical tools (i.e. normal probability plot, frequency histogram, residuals vs. ﬁt and vs. observation order plots), and the agreement of ﬁeld results with values predicted by this ﬁnal model was evaluated using linear regression analysis. The usefulness of the ﬁnal model was further evaluated using data from two other pond systems in Brazil: (i) pond system 1, an experimental pilotscale system rather similar to that of this study – a UASB reactor followed by a threematuration pond series (around 140 to 180 m2; 0.6–0.8 m deep; hydraulic loading rate of 0.15–0.22 m3/(m2.d); HRT of 1.5–4.3 d), situated in Belo Horizonte (near to Viçosa, with similar climate conditions); this system was monitored from January 2007 up to May 2009 and the dataset comprised 94 results for inﬂuent and efﬂuent ammonium concentrations (Assunção ); and (ii) pond system 2, an experimental system comprising an anaerobicfacultative pond series (full scale, 12 m3/d), followed by two maturation ponds in parallel (pilot scale, 16 m2; 0.5 and 1.0 m deep; ammonia SLR of 15–60 kg N/(ha.d)), situated in Lins, São Paulo State; this system was studied over June 2007 to December 2009, entailing 84 values for ammonia removal (Ruggeri ).
RESULTS AND DISCUSSION Equations (1)–(3) represent, respectively, the ‘complete’ multiple regression model, the stepwise model obtained with only the statistically signiﬁcant variables, and the ﬁnal model with only those variables which can be conﬁdently set at the designing stage. CeðNH3 Þ ¼ 62:8 þ 3835 SLRðNH3 Þ þ 0:022 BOD þ 0:928 HRT 30:8 h 0:382 T 0:0408 TSS
(1)
R ¼ 0:71
3:91 pH þ 0:00115 chl a
2
CeðNH3 Þ ¼ 65:727 þ 3826 SLRðNH3 Þ 31:1 h 4:04 pH þ 0:93 HRT 0:0371 TSS 0:42 T
R2 ¼ 0:71 (2)
CeðNH3 Þ ¼ 24:985 þ 3892 SLRðNH3 Þ 25:7 h 1:23 HRT 0:44 T
R2 ¼ 0:64
(3)
1899
Table 1
R. K. X. Bastos et al.


Water Science & Technology
Nitrogen removal in waste stabilization ponds


70.12
2014
Physical and operational characteristics of the pond system
Period 1 (October 2001–April 2002)
(1)
Period 2 (June–November 2002) (1)
Period 3 (March–May 2003)
Variable
P1
P2
P3
P4
P1
P2
P3
P4
P1
P2
Q
1.5
1.5
1.5
–
2.0
2.0
2.0
–
2.0
2.0
(2)
P3
P4
–
2.0
HRT
9.4
9.4
9.4
–
7.1
7.1
7.1
–
7.1
5.4
2.31
–
h
0.9
0.9
0.9
–
0.9
0.9
0.9
–
0.9
0.7
0.3
–
Q/A
0.093
0.093
0.093
–
0.123
0.123
0.123
–
0.123
0.123
0.123
–
Period 4 (March–September 2004)
(2) (3)
Period 5 (September 2004–August 2005)
(2) (3)
Period 6 (October 2005–March 2006)
Variable
P1
P2
P3
P4
P1
P2
P3
P4
P1
P2
Q
4.2
4.2
2.1
2.1
3.0
3.0
1.7
1.7
1.5
1.5
HRT
3.4
3.4
5.14
5.14
4.7
4.7
7.2
7.2
5.1
4.1
18.8
18.8
h
0.9
0.9
0.7
0.7
0.9
0.9
0.7
0.7
0.5
0.4
0.9
0.9
Q/A
0.259
0.259
0.130
0.130
0.185
0.185
0.093
0.093
0.093
0.093
0.046
0.046
Period 7 (April–August 2006)
(2) (4)
Period 8 (September–November 2006)
(2) (4)
P3
(2) (3)
P4
0.75
0.75
Period 9 (November 2006–February 2007)
Variable
P1
P2
P3
P4
P1
P2
P3
P4
P1
P2
Q
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
P3
(1) (4)
P4
1.5
1.5
HRT
5.1
4.1
4.1
9.4
7.2
7.2
4.1
4.1
7.2
7.2
4.1
4.1
h
0.5
0.4
0.4
0.9
0.7
0.7
0.4
0.4
0.7
0.7
0.4
0.4
Q/A
0.093
0.093
0.093
0.093
0.093
0.093
0.093
0.093
0.093
0.093
0.093
0.093
Period 10 (March–August 2007)
(2) (4)
Period 11 (October–December 2007)
(1) (4)
Period 12 (February–October 2009)
Variable
P1
P2
P3
P4
P1
P2
P3
P4
P1
P2
Q
1.5
1.5
1.5
1.5
2.5
2.5
2.5
2.5
3.5
3.5
3.5
3.5
HRT
9.4
9.4
9.4
9.4
5.6
5.6
5.6
5.6
4.0
4.0
4.0
4.0
h
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
Q/A
0.093
0.093
0.093
0.093
0.155
0.155
0.155
0.155
0.217
0.217
0.217
0.217
Period 13 (January–December 2010)
(1) (4)
Period 14 (January–December 2011) (1)
P4
(4)
Variable
P1
P2
P3
P4
P1
P2
P3
P4
Q
2.0
2.0
2.0
2.0
1.5
1.5
1.5
1.5
HRT
7.1
7.1
7.1
7.1
9.4
9.4
9.4
9.4
h
0.9
0.9
0.9
0.9
0.9
0.9
0.9
0.9
Q/A
0.123
0.123
0.123
0.123
0.093
0.093
0.093 3
P3
(1) (4)
0.093
2
Q: ﬂow rate (m³/d); HRT: hydraulic retention time (days); h: pond depth (m); Q/A: hydraulic loading rate (m /m d); P: ponds; (1) pond system fed with the UASB efﬂuent; (2) pond system fed with the BF efﬂuent; (3) pond 4 in parallel to pond 3; (4) ponds 3 and 4 in series.
where Ce(NH3) ¼ ammonium concentration in pond efﬂuent (mg/L); SLR(NH3) ¼ ammonium SLR (kg NNH3/ha. d); BOD ¼ biochemical oxygen demand in pond inﬂuent (mg/L); HRT ¼ hydraulic retention time (d); h ¼ pond depth (m); TSS ¼ total suspended solids in pond inﬂuent
(mg/L); T ¼ inpond middepth temperature ( C); pH ¼ inpond middepth pH; chl a ¼ chlorophyll a (mg/L). The statistical signiﬁcance of the complete model (Equation (1)) was indicated by analysis of variance (p < 0.001), warranting, therefore, further stepwise tests. The W
1900
R. K. X. Bastos et al.

Water Science & Technology
Nitrogen removal in waste stabilization ponds

70.12

2014
ﬁrst and the last ponds, respectively). Generally, the linear model overestimated ammonium efﬂuent values below approximately 15–20 mg/L in all ponds (thus, underestimating values above these ﬁgures), whereas the Pano and Middlebrooks model overestimated values as high as 40 and 30 mg/L respectively in the ﬁrst and second ponds, and below 15 and 5 mg/L in the third and fourth ponds, respectively. In short, both models did provide fairly good predictions for the more typical ammonium values of each pond in the series, with better agreements being found for intermediate values of ammonium concentration, i.e. in ponds 2 and 3, mainly for the linear model, whereas the Pano and Middlebrooks model did provide better estimates for high ammonium values in the ﬁrst two ponds and for low values in the two last ponds. Table 3 and Figure 2 show the results of similar analyses carried out between our ﬁeld data and values predicted by the Reed () and (Middlebrooks et al. ) models for TN removal. Overall, as with the Pano and Middlebrooks and the linear models for ammonium removal, the Reed model tended to overestimate low TN values, mainly in the ﬁrst pond, and underestimate high values, in a more marked way in the last two ponds. The values estimated by the Reed model matched the ﬁeld data around 30, 15 and 5 mg/L, respectively in the ﬁrst, second and in both the third and fourth ponds. A similar trend was observed for the Middlebrooks model, but it overestimated the measured values in a much more pronounced way than the Reed model. Table 4 and Figure 3, and Table 5 and Figure 4, show the results of the analyses carried out to evaluate the linear model derived in this work against ﬁeld data from two other pond systems in Brazil (pond systems 1 and 2 as described in the ‘Methods’ section). When tested against a pond system rather similar to that of this study (system 1),
subsequent parsimonious model (Equation (2)) includes cause variables (SLR and pH) for, and effect variables (TSS and pH) from, ammonium removal. It is noteworthy that this model is dependent on TSS, which is an indicator of algal biomass concentration in ponds, suggesting therefore that algal uptake may have played an important role in ammonium removal. However, this model remains dependent on variables which cannot be conﬁdently predicted beforehand (TSS and pH), being then of little practical use for predicting ammonium concentration in pond efﬂuents. Equation (3) in a way overcomes this limitation, was proved to be statistically valid (based on the graphical tests mentioned in the ‘Methods’ section, but not included herein) but suffers from other issues such as: (i) ammonium SLR was shown to be by far the input variable that most impacted the ammonium efﬂuent estimates; that is, although other input variables were statistically associated with the output, comparatively they had little effect in determining the ammonium concentration in the pond efﬂuent; and (ii) as detailed in the next paragraphs, the linear relationship does not seem to hold for the lowest and the highest values of inﬂuent ammonium concentration. Table 2 shows the results of the regression analyses carried out to verify the agreement of our ﬁeld data of ammonium concentration in pond efﬂuents with values predicted by the Pano and Middlebrooks model and by the linear model derived in this study (Equation (3)). Overall, ammonium concentrations predicted by both models were in reasonable agreement with ﬁeld data (Figure 1), even though they both tended to overestimate low values (noticeably the Pano and Middlebrooks model in the ﬁrst ponds) and, conversely, underestimate high values (more clearly in the last ponds) of ammonium concentrations (which are though not typical values in the
Table 2

Ammonium concentration in pond efﬂuents – relationships (regression equations and respective coefﬁcients of determination) between observed (ﬁeld data from this study) and estimated values by the linear model derived in this study and the Pano and Middlebrooks model
Linear model
Pano and Middlebrooks model
Pond Regression equation
R2
Regression equation
R2
1
NH3est ¼ 0.5702 NH3obs þ8.6037
0.36
NH3est ¼ 0.5662 NH3obs þ18.612
0.42
2
NH3est ¼ 0.6968 NH3obs þ5.5005
0.65
NH3est ¼ 0.7288 NH3obs þ8.9545
0.71
3
NH3est ¼ 0.7029 NH3obs þ5.8064
0.73
NH3est ¼ 0.8135 NH3obs þ3.051
0.70
4
NH3est ¼ 0.5834 NH3obs þ6.9665
0.45
NH3est ¼ 0.6456 NH3obs þ2.1333
0.59
Generala
NH3est ¼ 0.6358 NH3obs þ6.7011
0.64
NH3est ¼ 0.9219 NH3obs þ4.5126
0.76
a
All ponds together.
1901
Figure 1
R. K. X. Bastos et al.


Nitrogen removal in waste stabilization ponds
Water Science & Technology

70.12

2014
Ammonium concentrations in pond efﬂuents: observed values (ﬁeld data from this study) versus values predicted by the Pano and Middlebrooks model (right) and the linear model derived in this study (left). The dotted line represents the perfect ﬁt between the observed and the predicted values (y ¼ x), the continuous line represents the relationship between the observed and the predicted values obtained by linear regression.
1902
Table 3
R. K. X. Bastos et al.


Water Science & Technology
Nitrogen removal in waste stabilization ponds

70.12

2014
TN concentration in pond efﬂuents – relationships (regression equations and respective coefﬁcients of determination) between observed (ﬁeld data from this study) and predicted values by the Reed (1985) and Middlebrooks et al. (1999) models for TN removal Reed model
Middlebrooks model
Pond Regression equation
R2
Regression equation
R2
1
Ntotal est ¼ 0.527 Ntotal obs þ15.86
0.47
Ntotal est ¼ 0.678 Ntotal obs þ24.49
0.48
2
Ntotal est ¼ 0.6 Ntotal obs þ6.795
0.64
Ntotal est ¼ 0.918 Ntotal obs þ9.996
0.73
3
Ntotal est ¼ 0.599 Ntotal obs þ2.329
0.55
Ntotal est ¼ 0.94 Ntotal obs þ6.851
0.73
4
Ntotal est ¼ 0.528 Ntotal obs þ3.197
0.41
Ntotal est ¼ 0.84 Ntotal obs þ5.488
0.67
Figure 2

TN concentrations in pond efﬂuents: observed values (ﬁeld data from this study) versus values predicted by the models proposed by Reed (1985) (left) and Middlebrooks et al. (1999) (right). The dotted line represents the perfect ﬁt between the observed and the predicted values (y ¼ x), the continuous line represents the relationship between the observed and the predicted values obtained by linear regression.
1903
R. K. X. Bastos et al.

Table 4

Water Science & Technology
Nitrogen removal in waste stabilization ponds
Ammonium concentration in pond efﬂuents – relationships (regression equations and respective coefﬁcients of determination) between ﬁeld data from pond system 1 and estimated values by the linear model derived in this study
Pond
Regression equation
R2
1
NH3est ¼ 0.5773 NH3obs þ4.3127
0.60
2
NH3est ¼ 0.6718 NH3obs þ3.1875
0.63
3
NH3est ¼ 0.7206 NH3obs þ2.3632
0.76
NH3est ¼ 0.6279 NH3obs þ3.4737
0.71
a
General a
All ponds together.
the linear model derived herein well predicted the ﬁeld data, although, once again, somehow overestimating low values of ammonium concentration – up to approximately 10 mg/L, and underestimating values above this ﬁgure (Table 4 and Figure 3). However, when tested against pond system 2 the linear model worked reasonably well for predicting ammonium in the efﬂuent of the facultative pond, but not at all for the maturation ponds (Table 5 and Figure 4). Even in the case of the facultative pond, although the linear model provided good predictions for intermediate values of
Figure 3


70.12

2014
ammonium concentrations (around 10–15 mg/L), it now underestimated low values (even producing negative values) and overestimated higher values. In attempt to adapt the original Pano and Middlebrooks model to our experimental data, Equation (4) was derived based on ammonium and pH mean values (calculated for data subsets covering periods for which there was a more evident association between pH change and ammonium removal) and ﬁxing pH at 7 as the minimum observed value corresponding to which negligible ammonium removal was recorded (in fact during the whole study period pH never dropped below 7) (Figure 5). Even though the Pano and Middlebrooks variant model did not have a strong statistical basis (R 2 ¼ 0.45, Figure 4), it did predict ammonium efﬂuent concentrations in a reasonable way, although somehow augmenting the effect of overestimating the measured ammonium values in relation to the original Pano and Middlebrooks model (Table 6 and Figure 6). Ce ¼
Co 1 þ ½0:033167:(A=Q):e(0,669:(pH7))
(4)
Ammonium concentrations in pond efﬂuents: observed values (ﬁeld data from pond system 1) versus values predicted by the linear model derived in this study. The dotted line represents the perfect ﬁt between the observed and the predicted values (y ¼ x), the continuous line represents the relationship between the observed and the predicted values obtained by linear regression.
1904
Table 5
R. K. X. Bastos et al.


Water Science & Technology
Nitrogen removal in waste stabilization ponds
Ammonium concentration in pond efﬂuents – relationships (regression equations and respective coefﬁcients of determination) between ﬁeld data from pond system 2 and estimated values by the linear model derived in this study R2
Pond
Regression equation
Facultative
NH3est ¼ 1.5142 NH3obs – 6.0344 0.84
Maturation 1 (h ¼ 1 m)
NH3est ¼ 0.3422 NH3obs þ8.7899 0.51
Maturation 2 (h ¼ 0.5 m)
NH3est ¼ 0.2117 NH3obs þ14.85
0.11
CONCLUSIONS Applying regression analyses on a large ﬁeld dataset, a simple linear model depending only on design variables that can be previously and conﬁdently set (in other words, a pHindependent model), was derived and shown to predict ammonium efﬂuent concentrations of maturation ponds in agreement with our own ﬁeld data. However, when this same model was further evaluated using data from two other pond systems in Brazil, it still made good predictions of efﬂuent ammonium concentrations in a system rather similar to that of the present
Figure 4


70.12

2014
study, but did not work well for the other pond system. For this, and also because the linear function of the model does not seem to hold for both low and high values of ammonium (more clearly, below 5 mg L1 and above 40 mg L1), it is suggested that it should not be used as a pond design model. On the other hand, the modelling exercise presented herein does point out, within the prevailing climate and pond hydraulics and physical conditions tested in this work (such as HRT and pond depth), the outstanding effect of ammonium SLR in ammonium removal, which is in itself relevant information for design purposes, for instance for preliminary land area requirement evaluation. This study also adds further indication that, in general, the Pano & Middlebrooks () model and both the Middlebrooks et al. () and the Reed () models still make reasonable predictions of, respectively, ammonium and TN removal in maturation ponds, in spite of the suggestions gathered elsewhere that ammonia volatilization is not the main pathway for nitrogen and ammonium removal in WSP. However, it should be noted that all these models tended to somehow overestimate low ammonium efﬂuent
Ammonium concentrations in pond efﬂuents: observed values (ﬁeld data from pond system 2) versus values predicted by the linear model derived in this study. The dotted line represents the perfect ﬁt between the observed and the predicted values (y ¼ x), the continuous line represents the relationship between the observed and the predicted values obtained by linear regression.
1905
R. K. X. Bastos et al.

Water Science & Technology
Nitrogen removal in waste stabilization ponds
Table 6


70.12

2014
Ammonium concentration in pond efﬂuents – relationships (regression equations and respective coefﬁcients of determination) between observed (ﬁeld data) and predicted values by the modiﬁed Pano and Middlebrooks model
Figure 5

Pond
Regression equation
R2
1
NH3est ¼ 0.496 NH3obs þ15.941
0.42
2
NH3est ¼ 0.8461 NH3obs þ13.372
0.73
3
NH3est ¼ 1.1894 NH3obs þ6.6172
0.88
4
NH3est ¼ 0.5388 NH3obs þ2.2114
0.68
Linear regression between the coordinates Y [Ln ((C0 /Ce) 1)/(A/Q)] and X (pH  7) for the rearrangement of Pano and Middlebrooks’ original model (R 2 ¼ 0.45).
Figure 6

Ammonium concentrations in pond efﬂuents: observed values versus values predicted by the modiﬁed Pano and Middlebrooks model. The dotted line represents the perfect ﬁt between the observed and the predicted values (y ¼ x), the continuous line represents the relationship between the observed and the predicted values obtained by linear regression.
concentrations whereas underestimating higher ﬁeld data values. Hence, there certainly is need for further efforts, including more sophisticated approaches such as nonlinear or dynamic modelling but at the same time delivering practical design models, for predicting nitrogen removal in WSP.
ACKNOWLEDGMENTS The authors would like to acknowledge the Brazilian agencies FINEP, CAPES and CNPq for funding research projects based on which the database used here was developed and/or for providing student scholarships.
1906
R. K. X. Bastos et al.

Nitrogen removal in waste stabilization ponds
REFERENCES Assunção, F. A. L. Estudo de remoção de nitrogênio, com ênfase na volatilização de amônia, em lagoas de polimento de eﬂuentes de reatores UASB tratando esgotos urbanos de Belo Horizonte/MG (A nitrogen removal study, with emphasis on ammonia volatilization, in polishing ponds of UASB reactors efﬂuents treating domestic sewage in Belo Horizonte/MG). MSc dissertation, Department of Sanitary and Environmental Engineering, University of Minas Gerais, Belo Horizonte, Brazil. http://www.smarh.eng.ufmg.br/defesas/770M.PDF (accessed 22 April 2013). Assunção, F. A. L. & von Sperling, M. Importance of the ammonia volatilization rates in shallow maturation ponds treating UASB reactor efﬂuent. Water Science and Technology 66 (6), 1239–1246. Bastos, R. K. X., Rios, E. N., Dornelas, F. L., Assunção, F. A. L. & Nascimento, L. E. Ammonia and phosphorus removal in polishing ponds. A case study in southeast Brazil. Water Science and Technology 55 (11), 65–71. Camargo Valero, M. A. & Mara, D. D. Nitrogen removal via ammonia volatilization in maturation ponds. Water Science and Technology 55 (11), 87–92. Camargo Valero, M. A. & Mara, D. D. Ammonia volatilisation in waste stabilisation ponds: a cascade of misinterpretations? Water Science and Technology 61 (3), 555–561.
Water Science & Technology

70.12

2014
Crites, R. W., Middlebrooks, E. J. & Reed, S. C. Natural Wastewater Treatment Systems. CRC Press, Boca Raton, FL, USA. Middlebrooks, E. J., Reed, S. C., Pano, A. & Adams, V. D. Nitrogen removal in wastewater stabilization lagoons. Presented at 6th National Drinking Water and Wastewater Treatment Technology Transfer Workshop Kansas City, Missouri, August 2–4, 1999. http://www.bvsde.paho.org/ bvsacd/leeds/removal.pdf (accessed 17 May 2013). Pano, A. & Middlebrooks, E. J. Ammonia nitrogen removal in facultative ponds. Journal of the Water Pollution Control Federation 4 (54), 344–351. Reed, S. C. Nitrogen removal in wastewater stabilization ponds. Journal of the Water Pollution Control Federation 57 (1), 39–45. Ruggeri Jr., H. C. Póstratamento de eﬂuente de lagoa facultativa visando à remoção de nitrogênio amoniacal (Post treatment of facultative pond efﬂuent aiming at ammonia removal). PhD thesis, Polytechnic School, University of São Paulo, São Paulo, Brazil. http://www.teses.usp.br/teses/ disponiveis/3/3147/tde19072011103911/ptbr.php (accessed 17 May 2014). Silva, S. A., Oliveira, R., Soares, J., Mara, D. D. & Pearson, H. W. Nitrogen removal in pond systems with different conﬁgurations and geometries. Water Science and Technology 31 (12), 321–330. Soares, J., Silva, S. A., Oliveira, R., Araújo, A. L. C., Mara, D. D. & Pearson, H. W. Ammonia removal in a pilotscale WSP complex in northeast Brazil. Water Science and Technology 33 (7), 165–171.
First received 26 November 2013; accepted in revised form 11 August 2014. Available online 23 August 2014