Analysis of kinetics of local hydraulic losses on the laterals of radial wells at Belgrade groundwater source

Reduction of the yield of Belgrade groundwater source (BGS) radial wells is caused by the decrease in the area through which the water flows, due to clogging and narrowing of the pore space in the zone near the laterals and a decrease in the number and length of active laterals. This is due to two dominant causes: microbial colmation and corrosion (Dimkić et al. 2011a). The alluvial groundwaters are a valuable resource, as, for example, many pharmaceuticals and other organic compounds are eliminated at BGS through bank and aquifer filtration processes (Kovačević 2017; Kovačević et al. 2017), which eliminates the need for sophisticated treatment methods (for example Zinatloo-Ajabshir et al. 2020, 2021). The processes causing the reduction in yield, their correlation with selected parameters of groundwater quality, theoretical basis for defining the parameters that determine the hydraulic performance of the well and its temporal change, are presented in Dimkić et al. (2011a, 2011c) and Institute Jaroslav Černi (2010). Regression functions correlating the rate of increase of hydraulic losses to dissolved bivalent iron concentration and redox potential of groundwater are presented in Dimkić & Pušić (2014) and Institute Jaroslav Černi (2010).

Reduction of the number of laterals is a significant consequence of regenerations. The laterals are sealed off when they are physically damaged to the point when the aquifer material breaks into the interior of the lateral and closes it, often near the well shaft, which results in a drastic decrease in flow and large increase in the content of sand in the water.

In case that the material of which the laterals are made is prone to corrosion, and groundwater conditions are favorable, corrosion first causes an increase in the roughness of the material, resulting in large areas becoming more suitable for the development of microorganisms and the accumulation of their extracellular products and mineral deposits. This in turn causes the formation of numerous micro zones with conditions even more favorable for the corrosion, as well as the clogging of the pores in the zone around the laterals. The deterioration of the material of the laterals, combined with strong mechanical shocks during regenerations by jetting, eventually causes their destruction. Less than one third of all laterals has retained the structural integrity along the entire original length (Polomčić et al. 2016).

If the corrosion processes are not intense, the decrease in flow through the laterals which mechanical integrity is preserved, is caused by an increase in local resistances due to the clogging of pore space by accumulation of micro-organism products and their biomass. The dynamics of these processes can therefore be studied best by observing wells with laterals made of corrosion-resistant stainless steel. All laterals installed before 2005 at BGS were made of ordinary carbon steel. From that year on, stainless steel laterals were installed (using Preussag method) (Dimkić et al. 2011a). Since there are only few such wells, at present 8 (Rb-1, Rb-5 m, Rb-8 m, Rb-8, Rb-15, Rb-16, Rb-18 and Rb-20), out of which for five of them we have data (Rb-5 m, Rb-8, Rb-15, Rb-16 and Rb-20), some wells with older laterals were also selected to investigate the effects and causes of colmation (those for which it was assumed that the colmation dominated, and for which there are more than three data on local hydraulic losses during the period between 2006 and 2013). The number of active laterals, six or more active laterals in 2013, was used as a criterion by which the wells with dominant biochemical clogging were separated from other wells, a total of 29 wells.

Effects of well regenerations on hydraulic losses were quantitatively evaluated, and 41 periods of operation uninterrupted by regenerations were singled out and average values of flow per lateral, dissolved bivalent iron concentration and redox potential calculated for each period. The objectives were: (1) to obtain the best possible correlations between these parameters and the rates of colmation, (2) to develop a new mathematical model for calculation of drawdown in wells that includes the effects of regenerations, and the calculation of the increase of local hydraulic losses using the best of aforementioned correlations, and (3) to use this model as a base for simulation model implemented using Excel and macros in Visual Basic for Applications. The main result obtained by analyses using the simulation model was the quantification of impact of regenerations, and of the difference in the total amount of extracted water between the well operation at constant flow and at constant drawdown between the regenerations.

Figure 1(a) shows a typical flow pattern around the single well with horizontal laterals in the vertical section.

The maximum available drawdown ΔS_max is constrained by the static piezometer level and the minimum pump suction depth below the water level in the well shaft.

From 2006 to 2013, the flows in wells with horizontal laterals, groundwater levels in near piezometers, and the ones in the hinterland and between the wells, changes due to regeneration, as well as the physico-chemical quality of water (in the laboratory and in situ) were systematically monitored at BGS (Jaroslav Černi Institute 2010). Campaigns of well flow measuring and LHR values determining were conducted approximately two to three times a year, using a method based on a brief shutdown of the pumps (about 15 min) and monitoring the inflow of water into the well. This method was chosen as the wells were not evenly equipped with flow meters, and this assured that the measurements would be carried out in the same way at each well, with the satisfactory level of measurement accuracy. When measuring the capacity of wells, the parts of the groundwater source in which the measurements were performed were previously brought into a quasi-stationary state of exploitation by neighboring wells working continuously for at least 48 h. The level in the well was measured and recorded every second using logging divers, and at the same time monitored using a manual level meter. Based on the diameter of the well and the change in the water level in the well, the volume of water that entered the well in a time interval was obtained, which represents the value of the flow of a given well. Changes in groundwater levels in close piezometers were also monitored, and local drawdown ΔS calculated as the differences in levels between the well and the piezometer (Jaroslav Černi Institute 2010). In addition to hydraulic measurements, in situ measurements of physical and chemical parameters of water quality (temperature, standard hydrogen electrode redox potential, specific electrical conductivity, dissolved oxygen content, pH) were performed using probes. After measuring the physico-chemical characteristics, water was sampled from a well and a nearby piezometer for laboratory physico-chemical analyses of groundwater. Significant volume of data on the condition of the well and its change over time was obtained, as well as on the influence of physico-chemical parameters of groundwater quality on these changes.

The periods over which all 29 selected wells were monitored were divided into the sub-periods between regenerations, if there were any. Linear regression of LHR values was performed for each of the periods to obtain KLHR, and average values of flow per lateral, dissolved bivalent iron concentration and redox potential were calculated. Periods for which there was insufficient data were not taken into consideration. In this way, 41 periods with the corresponding average values of flow per lateral, bivalent iron concentration, redox potential and KLHR were isolated.

Previous research has shown that there is a relationship between KLHR and dissolved bivalent iron concentration, KLHR and flow per lateral or inlet water velocity, as well as KLHR and groundwater redox potential (Dimkić et al. 2008; Jaroslav Černi Institute 2010; Dimkić et al. 2011c; Dimkić & Pušić 2014). Therefore, the correlations between KLHR and average values of flow per lateral, dissolved bivalent iron concentration and redox potential for all wells, and all 41 periods of their operation, were examined. Functions of different types were examined, of which the best regression results were obtained for combinations of power and exponential function for the correlation of KLHR with the flow per lateral and concentration of dissolved bivalent iron, or KLHR with flow per lateral and redox potential. Correlation procedure was performed using Excel Solver in such a way that the values of the parameters ​​in the correlation function were automatically varied so that a minimum value was obtained in the target cell, which contained the mean squared error.

Evaluation of regeneration results included testing the correlation of y_reg with dissolved bivalent iron concentration, redox potential, number of regenerations performed so far on the well, LHR value before regeneration, number of active laterals before regeneration and flow per lateral, using basic forms of functions available within trend line option in Microsoft Excel.

The LHR at the end of each period is calculated as the sum of the resistance of the aquifer x, local hydraulic losses resulting from incrustations from previous periods (obtained by calculating the gradual growth of LHR depending on the corresponding KLHR in each previous period, and applying the coefficient of reduction of local hydraulic losses due to regeneration (y_reg) for all previous regenerations), and local hydraulic losses incurred during the observed period. Drawdown at the end of that period is obtained by multiplying LHR with the flow per lateral during the observed period (Equations (13)–(18)).

The maximum available drawdown ΔS_max is set by the static piezometric level and the minimum pump suction depth below the level in the well shaft. Maximum flows and maximum total amount of pumped water during each period are achieved by utilizing the total available drawdown, so it is assumed that the value of ΔS_max for each period is the same and equal to total available drawdown, assuming that the general goal is to use the well capacity as much as possible. The flow per lateral in between the regenerations is constant.

In practice, the operation of the well pump is usually regulated in such a way as to keep the level constant, hence the flow changes over time. Using the presented method, this case can be simulated by dividing the periods between regenerations (or, if there are none, the whole observed period) into a large number of intermediate periods within which the level does not change much even when the flows per lateral are large. In Excel Solver, it is not possible to vary the value of a sufficiently large number of variables (in this approach to the problem), in this case the value of y_n, i.e. the ratio between the flows per lateral in successive intermediate periods, and in addition the macro within which Solver would be used could only be used for a predefined number of periods. Also, the variation of y_n affects the calculated value of level drop through Equation (18), and it is hardly feasible for it to be developed in a single spreadsheet cell for a large number of periods, even by combining several of them.

Due to the mentioned limitations, alternative, less precise method using a macro was applied instead of the Solver – q_n was changed in steps of 0.01 starting from the value of 0.01 L/s and the drawdown at the end of period n calculated, until the absolute value of the difference between the calculated and set maximum drawdown starts to increase, which is a sign that in the previous iteration the difference was the smallest. In this way, the calculation could be performed for any number of periods, and much faster.

The results of the analysis of all periods between regenerations are given in Table 1 and Figure 2 (the graphs for all wells are available as supplementary material). The flows per lateral shown in the diagram in Figure 2 are the ones recorded when measuring levels, though there were usually more measurements in between, which were also taken into account when calculating average flow per lateral during a period without regeneration.

Dependence of KLHR on the concentration of dissolved bivalent iron and flow per lateral for 12 periods of operation of five wells with new, stainless steel laterals (Rb-5 m, Rb-8, Rb-15, Rb-16 and Rb-20) and the dominant biochemical colmation mechanism is shown in Figure 3.

Examination of the dependence of KLHR on the redox potential (standard hydrogen electrode) for 12 periods of operation of five wells with new, stainless steel laterals yielded the results of significantly worse quality and no sufficiently satisfactory correlation was achieved.

Because wells with older type of laterals have very large differences in permeability between laterals and along a lateral, it is difficult to make a connection between inlet velocities and flow per lateral, so no equation is presented that would correspond to Equation (25).

Regression analysis did not yield satisfactory results, and it can only be concluded with confidence from the distribution of KLHR values in relation to the redox potential that low KLHR values occur above 150 mV (matching well the observation in Dimkić et al. (2008) that the wells with values above 160 mV stand out with high yields, as well as the 150 mV limit adopted to label wells as being “good” or “bad” in Dimkić et al. (2011b), that between 85 and 150 mV KLHR values can range from 0.03 to 1.4 m/(L/s)/a, while below 85 mV all values are above 0.25 m/(L/s)/a.

During the 2006–2013 period, for which the necessary data are available (parameters significant for the effects of regenerations are given in Table 2), 27 regenerations were performed for which the value of y_reg could be determined, 13 of which were performed on wells belonging to the group with dominant biochemical colmation (italic font in Table 2), seven of which were the wells with new laterals (bold font in Table 2).

The effects of 27 regenerations (evaluated through the parameters significant for regeneration effects) performed during the period from 2006 to 2013, for which the necessary data are available, are presented in Table 2. Q₁ and Q₂ are well flows, q₁ and q₂ are flows per lateral, and q_1ave and q_2ave are the average flows per lateral during the periods before and after regeneration, respectively.

Of all the correlations examined, a weak linear correlation was found between flow per lateral before the regeneration and y_reg, and a slightly better power function correlation between local hydraulic losses before the regeneration and y_reg. The graphs are available as supplementary material.

Inlet velocities and flows per lateral with matching concentrations of dissolved bivalent iron for wells with new laterals (above) and laterals with dominant biochemical colmation (below) are shown in Figure 4. The dependence of critical velocities and critical flows per lateral on dissolved bivalent iron concentration according to Equations (31) and (32), with critical values from other papers, is also shown. In addition, velocities were singled out in those wells where the multiplication of the mean flow per lateral with KLHR resulted in an annual decrease in level of approximately 1.0 m. This is generally not an actual increase (as the level is usually kept close to minimum and the flow is slowly decreasing).

Regarding wells with dominant biochemical colmation, in Figure 4 (below), it can be observed that the zones of recommended velocities/flows per lateral according to Jaroslav Černi Institute (2010) match quite well the velocities/flows per lateral of 12 periods with an annual decrease in level of about 1.0 m (Table 1), and with the exponential function obtained by regression. The values of velocities and flows per lateral calculated by the exponential function obtained by regression for these 12 periods are somewhat higher than the critical values (for an annual decrease in level of 1.0 m) obtained on the basis of Equation (26). Recommended velocities/flows according to Dimkić & Pušić (2014) for ΔS_maxa = 0.35 m/a are in increasing percentage higher than the values obtained on the basis of Equation (26) below the dissolved iron concentration of 2.2 mg/L and in decreasing percentage lower above 2.2 mg/L. The differences are not surprising given that the correlation for KLHR and redox potential in that paper was derived based on the data from three other groundwater sources and one drainage system in the Danube, Tisza, and Velika Morava alluviums, and that the correlation with bivalent iron concentration was performed using a correlation between bivalent iron concentration and redox potential (Dimkić & Pušić 2014).

In wells with the new laterals, the critical velocities and flows per lateral obtained from Equations (31) and (32) for an annual level drop of about 1.0 m match the lower part of the zones of recommended velocities/flows per lateral according to Jaroslav Černi Institute (2010). The velocities and flows per lateral calculated using the exponential function obtained by regression (for only two periods in wells with new laterals having an annual level drop of about 1.0 m) are in increasing percentage higher than the critical velocities/flows obtained from Equations (31) and (32) above the dissolved iron concentration of 1.0 mg/L and in decreasing percentage lower below 1.0 mg/L (Figure 4).

Figure 5 shows a fairly good match between the values of calculated critical velocities and flows per lateral for a maximum annual level drop of 1 m with velocities and flows per lateral for 12 periods with an annual increase in level drop of about 1.0 m (Table 1). The match with the regression-obtained polynomial for these 12 wells (blue curve in Figure 5) is very good. Recommended velocities/flows per lateral according to Dimkić & Pušić (2014) for ΔS_maxa = 0.35 m/a (orange curve in Figure 5) are in increasing percentage higher than the values obtained on the basis of Equations (34) and (35) above 40 mV, which is not surprising considering that the correlation in that paper was performed on the basis of averaging the values over five ranges of redox potential, for a different set of wells at the Belgrade groundwater source and using additional vertical wells from the groundwater sources along Velika Morava river.

Upon comparison with the critical velocities and flows per lateral calculated on the basis of the concentration of bivalent iron, it can be noticed that the values are higher when calculated using redox potential. Since the quality of the regression is much better for correlations with bivalent iron concentration, it is advisable to use them, except for wells with high redox potentials, above 150 mV, as the KLHRs calculated using Equation (28) match data based on field measurements better in this range, than the ones calculated using Equations (24) and (26) at low concentrations of ferrous ions corresponding to high values of redox potential (Figure 3).

The decision on the maximum acceptable annual level drop depends also on the assessment of whether and how often regenerations will be performed, as well as the prognosis of their effects.

The simulation procedure was performed for the cases of 1, 4 and 7 regenerations for all combinations of KLHR₀, aquifer resistance x (drawdown per unit flow per lateral) and maximum available drawdown ΔS_max values, for each y_reg value. Using a macro written in the Visual Basic for Applications programming language, the calculations in Excel Solver are started automatically for all 4,851 combinations of number of regenerations (3 values), y_reg (11 values), KLHR₀ (7 values), x (7 values) and ΔS_max (3 values) and the results are recorded in separate spreadsheets (for cases with 1, 4 and 7 regenerations) and in one spreadsheet with complete results. The ratio of total volumes of pumped water during the observed period with (Q₁) and without regenerations (Q₀) was also calculated from the flows per lateral recorded for each combination of key parameters.

For one regeneration during the observed period total volume of pumped water during that period is 11% (y_reg = 1.0) tо 41% (y_reg = 0.0) larger than the total flow without regeneration (using Equation (33) – 0% (y_reg = 1.0) tо 59% (y_reg = 0.0)), for four regenerations 21% (y_reg = 1.0) to 123% (y_reg = 0.0) larger (using Equation (33) – from 21% (y_reg = 1.0) to 120% (y_reg = 0.0)), for seven regenerations 24% (y_reg = 1.0) to 182% (y_reg = 0.0) larger (using Equation (36) – 44% (y_reg = 1.0) to 164% (y_reg = 0.0)).

Based on Equation (35), the maximum allowable lateral flows and input velocities according to Equations (31), (32), (34) and (35) can be multiplied by approximately the square root of the number of regenerations increased by 1, to obtain the maximum allowable values in the first period of well operation. Based on the results of the calculation performed using a macro written in the Visual Basic for Applications and Excel Solver, in the already explained way, it was found that when the mean value of the coefficient of reduction of local hydraulic losses due to regeneration y_reg is about 0.5 ± 0.1, the lateral flow after all subsequent regenerations should be about 75% of that before the first regeneration.

Figure 6 (top) shows how the level inside the well changes at a constant flow between the regenerations. In practice, the operation of the well pump is usually regulated in such way as to keep the level constant, hence the flow changes over time (Figure 6 (bottom)).

Figure 6 (bottom) shows the change in flow per lateral during a 30-year period, for the case when there are no regenerations and the maximum drawdown is reached after 30 years (red line) of constant flow, and for the cases when pumping at maximum drawdown is simulated by dividing the 30 year period to 10, 25, 50, and 100 intermediate periods (violet, blue, green, and yellow curves, respectively). It can be observed that the initial, high flows can be obtained by calculation only if the intermediate periods are short enough (Figure 6). Approximately 30% higher maximum flows can be obtained in approximately twice as long periods. The shorter the intermediate period – the higher the flow and the calculated drawdown due to the aquifer losses and accumulated LHR, the lower the remaining available drawdown for the increase of local losses. At very short lengths of intermediate periods, level in the well practically does not change, and gradual decrease in flow at constant level is replicated in the calculation. The ratio between the total extracted water when pumping with a constant, maximum drawdown (simulated over 100 intermediate periods) and pumping at constant flow in such way that the maximum drawdown is reached after 30 years, Q_100p/Q, is equal to 1.39 for the particular combination of the resistance of aquifer and kinetics of local hydraulic losses (Figure 6).

The change in the flow per lateral during the 30 year well operation period is shown in Figures 7 and 8, in case there is no regeneration and the maximum drawdown is achieved after 30 years (red line), in case regenerations are performed (4 or 7 during the period in consideration) with maximum drawdown (10 m) achieved at the end of the period between regenerations (yellow line), and in the case of regenerations being performed and drawdown always equal to the maximum value (gray curve), all three cases simulated for different values of KLHR₀ and aquifer resistance x. The ratio between the total extracted water with drawdown constantly kept equal to the maximum between regenerations (simulated over 400 intermediate periods in total), and with maximum drawdown reached just before the next regeneration, marked as Q_400p/Q, ranges from 1.12 to 1.25 in the case of 4 regenerations, while in the case of 7 regenerations ranges from 1.10 to 1.18.

Pumping constantly at maximum drawdown according to the model produces some 20% larger volume of total extracted water.

By isolating wells in which colmation of laterals is a process that predominantly causes local hydraulic resistance increase, as well as by treating periods separated by regenerations as independent, satisfactory correlations between KLHR and dissolved bivalent iron concentration and flow per lateral were obtained (and to a certain extent useful correlation between KLHR and redox potential and flow per lateral).

Critical velocities and flows per lateral, given by Equations (31) and (32) in general match the values from previous papers and studies, mostly IJČ (2010), while critical velocities/flows according to Dimkić & Pušić (2014) for ΔS_max = 0.35 m/a for a dissolved iron concentration of 2.2 mg/L are the same as obtained by Equations (31) and (32), but they differ more and more as the concentrations get higher or lower than this value, which, having in mind that the average concentration for the analyzed periods is 1.55 mg/L, is not too large a deviation.

The impact of regenerations on the total amount of pumped water is large. In the case of maximum analyzed number of 7 regenerations (approximately once in 4 years for a period of 30 years) at y_reg = 0.4, which is the average value for analyzed regenerations (Table 2) at Belgrade groundwater source, it is two times larger than in the case of no regeneration. The correlation connecting the ratio of the amount of pumped water with (Q₁) and without regeneration (Q₀) and the resistance of the aquifer x, the coefficient of reduction of local hydraulic losses due to regeneration y_reg and the number of regenerations n is given by Equation (36). According to Equation (38), the maximum recommended values of velocities and flows per lateral (obtained from Equations (31), (32), (34) and (35)) at the beginning of well operation can be multiplied by approximately the square root of the number of regenerations increased by 1. After regeneration, the flow per lateral should be about 75% of that before the first regeneration if y_reg is about 0.5 ± 0.1. The effects of regeneration cannot be predicted, only weak correlations have been established between flow per lateral before regeneration and y_reg, as well as between local hydraulic losses before regeneration and y_reg, meaning that the effects of regeneration increase moderately with the magnitude of local resistances. The approach that can be proposed is to use the y_reg value for design calculations by 0.1–0.2 higher than the average for the groundwater source.

In practice, the well pump is operated in such a way that the level is kept constant and the flow decreases over time. This mode of operation is simulated by dividing the period between regenerations into a large number of smaller periods in which the flow is constant, and it changes between these periods in such a way that the level at the end of the period is the same as at the end of the previous one. In this way, the phenomenon of high initial flows, which decline very quickly, is successfully replicated (Figures 6–8). The ratio between the volume of total extracted water for the case when the drawdown is constantly kept equal to maximum (simulated over 400 intermediate periods) and when the flow between the regenerations is kept constant (as the level in the well decreases), is between 1.12 and 1.25 for 4 regenerations, while for the case of 7 regenerations is between 1.10 and 1.18.

Providing that the effects of regeneration are the same, after several regenerations the calculations show that the flow does not decrease further as LHR has become large enough that the percentage eliminated by regeneration becomes equal to the resistance accumulated during the period after the previous regeneration.

The developed software model can be used to simulate different scenarios for number of laterals, well operation mode and regeneration, or Equations (31), (36) and (38) can be used, together with ratios that pertain to the differences between the constant level and constant flow mode, to predict the well yields and total extracted volumes.

The authors would like to thank the Ministry of Education, Science and Technological Development of the Republic of Serbia for the financial and other kinds of support within the project ‘Methodology for Assessment, Design and Maintenance of Groundwater Sources in Alluvial Environments Depending on the Aerobic State’, Project No. TR37014.

All relevant data are included in the paper or its Supplementary Information.