*3.1. E*ff*ects of the Hydrocyclone Filtration and Resting Time on the Traits of the Digestate Liquid Fraction*

Filtration of the digestate liquid fraction (DLF) influenced the pH of the different resulting fractions (Figure 2) pointing out that the native fraction before filtering had a pH value higher than 8.2 and not statistically appreciable differences were found after the filtration (hereafter referred as hydrocyclone filtered DLF, or HF-DLF). When the HF-DLF was allowed to rest for eight days before the beginning of the experiment, the solution at 50% dilution showed a lower pH compared to both the freshly made native DLF and the freshly made HF-DLF soon after the filtration. However, such latter difference was not statistically appreciable according to the conservative post-hoc test used. The pH value of the native DLF was higher than the fraction discarded from the filter and the tap water.

**Figure 2.** Values of pH of the digestate liquid fraction (DLF), before and after hydrocyclone filtering (HF), the 50% dilution of the HF-DLF, and the tap water used for the experiment. Bares are least square means (LSmeans) ± Lsmeans standard error estimates. Bars with a letter in common cannot be considered different according to a conservative Tukey-grouping applied to the *p*-differences of the LSmeans. Results of the statistical analysis are embedded.

The output of the turbidity sensor did not change by the filtration among the DLFs (Figure 3), which recorded value closed to 14.2 mV (please mind that, in theory, the lower the conductibility, the higher the turbidity). The tap water showed a turbidity sensor output close to 800 mV, 56-fold than any of DLFs or HF-DLF, on average.

**Figure 3.** Output of the turbidity sensor in the digestate liquid fraction (DLF), before and after hydrocyclone filtering (HF), the 50% dilution of the HF-DLF, and the tap water used for the experiment. Bars are LSmeans ± Lsmeans standard error estimates. Bars with a letter in common cannot be considered different according to a conservative Tukey-grouping applied to the *p*-differences of the LSmeans. Results of the statistical analysis are embedded.

The dry matter concentration varied among the DLFs (Figure 4), with the fraction discarded from the filter showing a relative concentration compared to the native DLF 45% higher. The native DLF also showed a marginally, albeit significantly, lower dry matter concentration than the HF-DLF (−1.1% relative difference, corresponding to −0.02%).

**Figure 4.** Dry matter concentration of the digestate liquid fraction (DLF), before and after hydrocyclone filtering (HF), the 50% dilution of the HF-DLF, and the tap water used for the experiment. Bars are LSmeans ± Lsmeans standard error estimates. Bars with a letter in common cannot be considered different according to a conservative Tukey-grouping applied to the *p*-differences of the LSmeans. Results of the statistical analysis are embedded.

## *3.2. E*ff*ects of the Digestate Liquid Fraction Dilution on the Emitter Performances and Solution Traits*

Results of the statistical analyses of the irrigation test are shown in Table 2. Both the amount of water in the control and HF-DLF and its turbidity varied by the treatment at increasing time, with differences more marked among treatments in the early stages of the irrigation. The dilution of the DLF influenced the quantity of the solution released by the emitters during the test depending on the percentage of the dilution (Figure 5, Supplementary Material Table S1). In particular, water release increased almost constantly, whereas 10% and 25% dilution during the first 3 h. The 50% showed milder increases, and a total amount of water released slightly lower than the other treatments. The coefficient of variation of the system was in general lower than 5%, with some outlier only in the HF-DLF 25% dilution (Table 3).

**Figure 5.** Amount of solution (tap water in the control and digestate liquid fractions (HF-DLF) diluted at the 10%, 25%, and 50%) released by the emitter each hour (upper left panel); dry matter concentration of the DLF released (upper right panel); turbidity of the HF-DLF released (lower left panel); and amount of OH<sup>−</sup> ions released per ton of water released each hour. Data are LSmeans ± LSmeans standard error estimates. For post-hoc comparisons and raw data see Supplementary Material Table S3.


**Table 2.** Results of the statistical analysis (degrees of freedom estimate of the error (DF den); F statistics; and *p* values) of the general linear mixed model appliedtheamountofsolutionreleasedeachhour,itsturbidity,drymatterconcentration,pHdifferencecomparedtothecontrol,andtheamountofOH−releasedwith

 to

6

7

8

 111.8

 2.5

 0.065

 95.4

9.5× 103

<0.0001

15.5

 112.9

<0.0001

25.5

 226.2

<0.0001

20.5

 119.0

<0.0001

 111.7

 4.4

 0.006

 95.4

9.5× 103

<0.0001

20.3

 159.0

<0.0001

25.5

 183.6

<0.0001

20.5

 110.3

<0.0001

 111.6

 3.0

 0.034

 95.4

9.8× 103

<0.0001

13.2

 142.3

<0.0001

25.5

 205.9

<0.0001

20.5

 140.0

<0.0001


**Table 3.** Coefficient of variation of each treatment (n = 9, consisting of 3 lines with 3 emitters each) and relative distribution percentiles at 0.025, 0.25, 0.5 (median), 0.75, and 0.975. Data of the three dilutions of the hydrocyclone filtered digestate liquid fraction (HF-DLF) were showed singly and bulked.

The dry matter content of the solutions (Figure 5) showed that the concentration strongly depended on the dilution rate, and to a lesser extent on time. In fact, higher values >0.6%) were recorded in the DLF 50%, compared to those in DLF 10% < 0.2%) and DLF 25% showed values ranging from 0.4% to 0.3%. The value of dry matter concentration (%) was quite stable all along the 8 h of irrigation test for all the treatments, except for DLF50% that showed a slight increase with time (Supplementary Material Table S1; Supplementary Material Table S2).

A similar trend, but more pronounced by the time, was found regarding turbidity (Figure 5; please mind that the higher is the turbidity value, the lower the liquid turbidity). For all the treatments the values recorded were stable over time and around 60 mV, 160 mV, and 380 mV for DLF 50%, 25% and 10% respectively. The analysis of the concentration of ions OH− (Figure 5), calculated using the pH values, showed that even if the trend of treatments DLF 10% and 25% was slightly variable during the irrigation test, the values were included between 5 and 10 OH<sup>−</sup> mol t−<sup>1</sup> solution, with scarce differences by the time. Instead, treatment DLF 50% showed an increasing trend during the test with an initial value of 13 OH<sup>−</sup> mol t<sup>−</sup>1and a final value of 25 OH<sup>−</sup> mol t<sup>−</sup>1. Finally, we inspected a serpentine from the control and the DLF 50% (Figure 6) and found that no clogging occurred.

**Figure 6.** Serpentine inspection of the control (tap water; above) vs. 50% diluted hydrocyclone filtered digestate liquid fraction (below) emitters.

### **4. Discussion**

In the present work, we studied the role of an increasing ratio between a hydrocyclone-filtered digestate liquid fraction (referred to as HF-DLF) and tap water on the performance of an irrigation system and water quality. Treatments included 3 HF-DLF ratios (10%, 25%, and 50% of total solution used for the irrigation) in contrast to tap water as control and measurements were taken at an hourly basis on an 8-h irrigation cycle, that simulates most of the irrigation cycles occurring in a broad range of crops.

The native digestate liquid fraction used before the hydrocyclone filtering had a higher pH than the tap water used (pH = 8.23 vs. 7.84, respectively) and such pH slightly reduced after the hydrocyclone filtering. Information on the effect that the hydrocyclone filtering has on the pH of digestate liquid fraction and its total solids are scarce, nonetheless, centrifuge filtering was shown to have relatively high efficiency [25]. Thus, the pH reduction after the hydrocyclone treatment may have been due to the fractionation of the calcium carbonates or other high-weight solids in the digestate liquid fraction, including cations. The digester diet of the material in the present study was mainly composed of corn and barley biomasses (residual and dedicated) and to a lesser extent of cow slurry. This kind of digestate has high contents in potassium, chloride, carbonates, and proteins [20]. It is thus likely that the high pH of the HF-DLF under study was due to a high content of basic, high molecular weight proteins, which can be removed by hydrocyclone filtering. Such a hypothesis is corroborated by the further reduction of pH of the DLF found eight days after filtering, before the irrigation experiment started, which may have been due to oxidation of the organic material in the HF-DLF that in such time-lapse was resting. Hydrocyclone filtering, however, did not result in a reduction of the turbidity and such results could be due to the high total solid concentration in the DLF following incomplete filtering, as pointed by Guilayn et al. [25]. Indeed, in our study, the HF-DLF showed a dry matter concentration of 16.0‰ (on a weight basis) and a pH = 8.15 soon after the filtration. These traits suggested a low quality fraction for drip irrigators according to the early classification by Nakayama and Bucks [3]. This likely was a main cause of the differences among the amount of HF-DLF released by the emitters at increasing time and varying the HF-DLF ratios within the irrigation system.

The manufacturer declared the used emitters as and releasing 2 L h−<sup>1</sup> at 100 kPa. When subjected to the 200 kPa pressure of the present study, we found that the amount of water released in the control ranged from 3.07 L in the first hour and such an amount increased on average by the 1.9% h <sup>−</sup>1. Such variation were higher than those found by Bodole et al. [26]. The variation of the amount of HF-DLF each emitter released per hour also increased with time in the three dilution treatments (0.5–0.9% h<sup>−</sup>1), but to a lesser extent compared to the control. Such variations likely depended on the usury of the system. In particular, the dilution at 10% and 25% released 1.9% and 3.5%, respectively, more HF-DLF per cycle than the water released by the control, whereas the dilution at 50% released 4.9% less HF-DLF per cycle than the control. Besides, the differences between each HF-DLF dilution and control in the amount of water released per unit time declined with time. The temperature of the tap water or the HF-DLFs was similar among the treatments and increased linearly during each irrigation cycle, starting at 9.42 ◦C ± 0.27 ◦C at the 9:00 a.m. (moment of the beginning of each experiment) and increased at a rate of 0.94 ◦C h−<sup>1</sup> <sup>±</sup> 0.05 ◦C h−<sup>1</sup> (data not shown). This implies that differences by time can only partly be explained by a heating of the emitters and thus the expansion of their pore size. We hypothesize that the emitters used in this experiment rapidly lost their ability to compensate for the irrigation rate at the pressure we used. In addition, the 10% and 25% diluted HF-DLF did not likely consist in a strong occlusion of the emitters. This is consistent with the constant rate of the dry matter content of 10% and 25% HF-DLF and increasing rate of the dry matter content of the 50% HF-DLF, which progressively increased at a rate of 27.98 <sup>×</sup> 10−<sup>3</sup> <sup>±</sup> 0.95 <sup>×</sup> 10−<sup>3</sup> pH units h−1. When using the 50% dilution, 6.63% less HF-DLF was released, on average, if compared to the water released in the control or the HF-DLF in the 10% and 25% dilutions. Despite such difference, the potential effect on the pH (expressed as [OH−] amount of a putative medium receiving the HF-DLF at the 50% dilution) strongly increased over time, whereas it did not vary in the in the control or the HF-DLF in the 10% and 25% dilutions

1.68 <sup>±</sup> 0.04 Mol OH<sup>−</sup> (t HF-DLF)−<sup>1</sup> h<sup>−</sup>1. Results from other experiments were variable and depended on the pressure, kind of emitter and kind of wastewater. In contrast to our study, Gamri et al. [27] found a strong reduction of the emitter performance with time and difference between the present and the one by Gamri et al. [27] experiment can be due to the higher pressure we used, which is two-fold compared to their study, and this occurred despite our HF-DLF had a solid concentration 2.3–9.8 fold higher than the synthetic wastewater composition used by Gamri et al. [27]. Nonetheless, these differences can be due to the high-frequency flushing in our experiment (one every 8 h). And indeed, Puig-Bargués et al. [14] showed that flushing every 540 h was sufficient to almost completely avoid the emitter clogging. Puig-Bargués et al. [14] also found, as in the present study, that dripline flow increased 8% and 25% over time when using a pressure compensating and a non-pressure compensating emitter, respectively, when used with a tertiary effluent from a wastewater treatment plant filtered with a 0.130 mm filtration level. Similarly, we found that coefficient of variation computed at an hourly basis was 1.653% (CI95% 1.240–2.065%) in the control and 3.587% (CI95% 2.713–6.444%) in the HF-DLF, with scarce differences among dilution, suggesting that accumulation of deposited material in the emitters affected the dripline flow performance. Such coefficient of variation was lower than those found in other similar studies [28,29] and can be marginally acceptable as indicated by Bodole et al. [26], according to which a test duration of more than 60 min is enough to minimize the uncertainty due to the initial fluctuation of the data. The lower variation of the HF-DLF is likely due to an anti-clogging shape of the present emitters if compared to other emitters [30].

## **5. Conclusions**

In conclusions, hydrocyclone filtration scarcely affected the traits of the digestate liquid fraction used for the irrigation. Irrigation with hydrocyclone-filtered digestate liquid fraction (HF-DLF) injected in the system at 10% and 25% dilution did not affect the performance of the system nor the traits of the liquid fraction released by the emitters, whereas using 50% dilution of the HF-DLF consisted in a lower amount of liquid released at increasing pH. In particular, HF-DLF dilution at 10%, 25%, and 50% consisted in +1.9%, +3.5, and −4.9% amount of liquid released compared to the control. In 10% HF-DLF % and 25% HF-DLF, a constant pH difference of + 0.321 ± 0.014 pH units compared to the control was found, in the 50% HF-DLF pH increased by around a half point and such difference increased with time.

This implies that that highly concentrated digestate liquid fractions, i.e., low dilutions, can pose problems for the functioning of the system and may have potentially harmful effects on soils with high pH. Nonetheless, the use of digestate liquid fractions for irrigation purposes may be a valuable option in those areas with a high amount of biogas plants and digestate production, such as various nations in Europe, America and Asia including USA, China, Germany, United Kingdom, Italy, and France [31,32]. Results from the present study have beneficial implication on the on water conservation since digestate production by feeding the digester with barley and corn provide wastewater from the late spring to the early fall, when water requirements are high. At the one time, the ability to use such wastewaters with minimal impact on the irrigation system, and thus with reduced negative impacts due to the system maintenance and disposal.

The digestate liquid fraction used in the present study was previously subjected to an additional hydrocyclone filtering, that likely discarded the high-molecular weight fraction. However, since few differences were found between filtered and non-filtered liquid digestate fraction, it is likely that a dilution at least up to 25%, according to the present study, can allow for a direct use of the digestate liquid fraction in microirrigation system with a minimal harming of the system performances. However, since the present is a short-term experiment, these results would require additional experiments with unfiltered liquid fractions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4395/10/8/1150/s1, Table S1: LSmeans estimates and relative standard errors of the traits under study [Solution released; output of the turbidity sensor; dry matter concentration; [OH−] concentration; and pH difference than control]. LSmeans

with a letter in common can't be considered different according to a conservative Tukey-grouping applied to the p-differences of the LSmeans.; Table S2: Linear models of the variation in time of the temperature in the whole experiment and of dry matter concentration, [OH−] concentration and pH, and in the 50% diluted hydrocyclone filtered digestate liqud fraction; Table S3: Raw data of the amount of water or diluted hydrocyclone filtered digestate liqud fraction, the output of the turbidity sensor (mV), dry matter content (%), pH of the solution, and the amount of [OH−] concentration.

**Author Contributions:** Conceptualization, S.B., M.B., E.R., S.S., L.P.; methodology, S.B., M.B., E.R., S.S.; software, S.B., E.R., S.S.; validation, S.B., S.S.; formal analysis, S.S.; investigation, S.B., M.B., E.R., S.S., P.C.; resources, M.C., P.T., C.B., L.P.; data curation, S.B., S.S., P.C.; writing—original draft, S.B., S.S.; writing—review and editing, S.B., S.S., P.C.; supervision, L.P.; project administration, L.P.; funding acquisition, L.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was carried out within the AGROENER project (D.D. n. 26329, 1 April 2016) funded by the Italian Ministry of Agriculture (MiPAAF).

**Acknowledgments:** The authors wish to thank Ivan Carminati, Gianluigi Rozzoni, Alex Filisetti and Elia Premoli for their assistance in performing the tests and for their professionalism and availability.

**Conflicts of Interest:** The authors have no conflicts of interest to disclose. The authors declare that they have no financial interests or personal relationships with the brands cited, nor endorse any of the brands cited or their products. Furthermore, the funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
