Experimental study on the operational efficiency of modified sandwich filters for pesticides and phosphate removal from tile drainage water

Abstract

Pilot aquifer storage and recovery (ASR) systems are currently being built and operated to capture agricultural run-off water during wet seasons in an attempt to increase the availability of water during the dry seasons. The collected water requires treatment prior to infiltration due to its high fertilizers and pesticides concentrations in order to adhere to Dutch legislation surrounding underground storage. This study investigated the performance of two different carbon types which are currently in use in a pilot ASR plant in Texel, which used slow sand granular activated carbon (SSF-GAC) sandwich filters. The GACs used in this study, one mesoporous, Eversorb 520 (GAC-E), one microporous, Norrit PK1 (GAC-N), where compared through isotherm experiments and lab scale SSF-GAC sandwich filters, constructed for the first time in a way to allow for sampling in between layers, during a 14 week period. Additionally, one sandwich filter was augmented for the first time with an Iron Oxide Coated Sand (IOCS) top-up layer to asses its ability to remove phosphate and natural organic matter (NOM) from agricultural water. The water was doped with 10 μg/L for 5 pesticides commonly found in Dutch agricultural water: Atrazine, Bentazone, Chloridazon, Imidacloprid and Tebuconazole. Of these compounds, Bentazone showed extremely weak adsorption during all studies. It is unclear if the weak adsorption to both carbon types is contained to Bentazone as a compound or due to its positive charge. The isotherm experiments in high NOM water (C0 = 13.85 mg/L) resulted in similar NOM loading on both carbon types, despite their difference in pore size distribution. The adsorption of pesticides was favored by GAC-E over GAC-N with higher loading (40 – 300 % in qe) where it appeared that the surface functional groups of the GAC where the dominant factor in the difference in adsorption. Isotherm studies with the microporous GAC-N in two water types with varying NOM concentrations (C0 = 13.85 mg/L vs 23.35 mg/L) found limited reduction in sorption capacity (30 – 40 % in qe ) for all compounds despite the fact that NOM loading on the GAC increased by 117 %.
During the column experiments the NOM loading was higher for the microporous carbon in contrast with the isotherm experiments, despite equal performance of the SSF in both columns. The adsorption of pesticides (EBCT 11.2 min) showed similar correlations as during the isotherm experiments, with 20 - 35 % higher breakthrough observed for GAC-N. Compensated for carbon density, the overall loading (μg/gGAC) was on average 45 % higher for GAC-N ( 𝜌=250𝑘𝑔/𝑚3 ) than GAC-E( 𝜌=500𝑘𝑔/𝑚3), contradicting the isotherm experiments. Due to problems with gas accumulation between the GAC in combination with wall effects the empty bed contact time (EBCT) was negatively impacted, resulting in a mass transfer zone (MTZ) that was too short for the compounds to reach equilibrium over the columns. If GAC-N has higher adsorption kinetics than GAC- E it could explain the overall higher pesticide loading. The reduction in EBCT resulted in immediate breakthrough from week 1 onwards, with breakthrough curves showing linear patterns. As a result, it was not possible to accurately predict the sorption capacity of the columns or translate lab-scale performance to pilot data. Modelling with fixed bed adsorption software using homogenous surface mass diffusion (HSMD) and linear driving force (LDF) models was attempted but did not yield usable data for all pesticides.
The IOCS isotherms (C0 = 50 mg PO43-/L) showed Freundlich type adsorption of phosphate (𝑘𝑓 = 5.39 & 𝑛=2.04), with additional high removal of calcium (37%), magnesium (27%) and potassium (10%), though these did not show Freundlich or Langmuir type adsorption. Phosphate removal was successfully modelled with HSDM models which predicted breakthrough at 700 - 2000 bed volumes (BV) depending on the diffusion coefficients. The column experiments showed significantly faster initial breakthrough at 400 BV despite maintaining 55 min EBCT. The column maintained a significantly higher sorption capacity after initial breakthrough for longer than predicted, losing only 50% capacity over the course 2500 BV. During the column studies high removal of calcium was observed and it was theorized that phosphate formed calcium precipitation complexes at pH of 8. The removal of magnesium and potassium was absent during column studies. At the conclusion of the experiments, the IOCS had sorbed 12.8 mg PO43 /gIOCS after 2573 BV with roughly 50% residual sorption capacity remaining.
As a result of pre-loading during production, the IOCS leached NOM during the isotherm experiments. Subsequent column experiments did not show NOM (UV254) desorption but rather showed NOM removal at higher rates than the SSF (10% vs 5%) which resulted in additional higher removal of NOM in the following GAC-E layer vs a column without the IOCS layer, exhibiting synergies between the layers.
The IOCS layer was successful in the removal of Imidacloprid from the influent, which started after 1100 BV or 6 weeks and is suspected to occur through biodegradation, peaking at 70 % removal at conclusion of the experiments. The SSF layer following the IOCS was presumably inoculated with the biomass and additionally removed 70 % of the Imidacloprid from the IOCS effluent. The combined IOCS-SSF removed 90 % of the Imidacloprid influent (μg/L), which has not been seen in these filters at such short EBCT to date for any compound. The SSF-GAC columns that did not contain IOCS did not show Imidacloprid removal, indicating that the biomass can only form on selective substrate. It is unclear at this point whether removal is contained to Imidacloprid or if additional compounds are susceptible for removal in IOCS layers. The combined influence of lower NOM and lower total pesticide loading on the GAC-E layer resulted in a 8 – 10 % increase in total pesticide adsorption vs the column that did not contain an IOCS layer, with 5 – 10 % lower breakthrough for all compounds. Chloridazon and Tebuconazole where likewise removed after 1100 BV or 6 weeks through suspected biodegradation in both the IOCS and SSF layers, peaking at 25 % and 40 % removal respectively. Though the biodegradation of these compounds has been proven in literature, it has not been observed in column studies at such high concentrations & removal rates at these low EBCT. It is hypothesized that the abundance of nutrients allowed for rapid bio growth and subsequent pesticide degradation.
The results of this study indicate that the augmentation of SSF-GAC sandwich filters with IOCS columns aid in the removal of phosphate, NOM and Imidacloprid, thereby extending the filters bed life and improving overall performance. While the removal of other pesticides remains to be investigated, the findings of this thesis underpin the use case of IOCS as a top up layer for SSF-GAC sandwich filters used to treat agricultural waters.