Twenty maple producers in Vermont were recruited to participate in the study during the 2009 production season, 5 each that typically concentrated their sap in the following 4 ranges: 8% - 11%, 12% - 15%, 16% - 20%, and 21% - 25% sugar, respectively. The objective was not to provide a representative sample of membrane types, configurations, or RO equipment, but simply a snapshot of the common ranges of concentration that producers were typically using at the time. Prior to the production season, the producers were provided with instructions and a set of glass sample collection bottles. They were instructed to collect samples of raw sap (“sap”), and the concentrated sap (“concentrate”) and water (“permeate”) produced from that sap at 3 times during the production season, at timepoints representative of the early, middle, and late portions of their production seasons. The duration and timing of the maple production season varies annually and between individual operations due to differences in climate and operational factors. As such, producers were instructed to collect these samples at times representative of the first, middle, and last thirds of the production seasons of their individual operations. Producers were instructed to collect the set of sap, concentrate, and permeate samples simultaneously, and to place them in a freezer immediately. At the end of the production season, all samples were collected from each producer and placed in a freezer at the University of Vermont Proctor Maple Research Center (PMRC) in Underhill, Vermont. Analyses to determine the physiochemical properties and chemical composition of the samples were conducted during the summer of 2009.

Several samples were lost prior to analyses due a variety of factors such as producer freezer malfunction or spoilage prior to transport to PMRC. Thus, there are several producers for which there is not a complete set of data from early, middle, and late timepoints, and all samples from some producers were lost. Ultimately, 34 sets of sap, concentrate and permeate samples from 14 producers were analyzed and included in the analyses.

Table 1 presents the overall mean, minimum, and maximum composition and properties for all sap, concentrate, and permeate samples. Average concentration levels of the concentrate samples from the 14 producers ranged from 8.9 to 19.8 ˚Brix, with average pH values slightly below neutral, 6.96, and ranging from just below to slightly above neutral (Table 1). Further interpretation of overall mean composition values for the concentrate samples is not possible due to the wide

range of concentration levels of the samples analyzed. However, the overall means for sap and permeate samples provide a portrait of the average composition of sap and permeate samples from a variety of producers across a single production season.

The sugar concentration of sap ranged from 1.5 to 2.4 ˚Brix, with an average of 2.0 ˚Brix (Table 1). pH was near neutral to slightly acidic, ranging from 6.10 to 7.40, with an average of 6.61 across all producers and time of season. Sucrose was the predominant carbohydrate, with trace amounts of glucose and fructose also present (Table 1). The minerals found in the highest concentrations were K and Ca, with mean concentrations of 64.6 and 49.1 ppm, respectively. The minimum, maximum, and standard error values for the concentrations of all minerals analyzed illustrate that there is relatively substantial variability in the mineral composition of sap (Table 1). All of these observations are consistent with the available published composition data for maple sap [1] [9] [10] .

Permeate samples were more acidic than sap, with an average value of 5.86, and ranging from 5.41 to 6.36 (Table 1). This is consistent with the properties of membranes used in separation processes. Because membranes are permeable to carbon dioxide gas but not to the bicarbonate ions which typically balance the effects of CO₂ on pH, the pH of the concentrated material will typically be slightly higher, and that of the permeate water slightly lower, than the incoming material being concentrated. This expected pattern can likewise be observed in the concentrate samples, which had an average pH slightly higher than that of the sap samples (Table 1). The average conductivity of permeate samples was low, 15.1 μS/cm, well below the generally recommended maximum of 50 μS/cm [12] , indicating that the membranes used by the RO machines included in this study were generally not allowing a large quantity of mineral ions to pass through into permeate. Likewise, the average ˚Brix of permeate samples was also low, 0.1%, indicating that the membranes of RO units in this study were also not allowing significant amounts of sucrose to pass through and be lost in permeate.

Table 2 presents the overall means for the sample sets from each producer, which can provide information not discernible from the overall means. Only 1 set of samples had conductivity values in permeate greater than the maximum recommended threshold (Producer A, Table 2). Mineral analysis of these samples confirmed that the conductivity was likely elevated by the presence of mineral ions, in particular Ca, K, Na, and Mn, indicating that the membranes of this RO unit were likely not functioning optimally, and were allowing elevated levels of mineral ions to pass through into the permeate. Because of this, the data from this producer’s samples were not included in any other analysis (e.g. Table 1, Table 3, and Table 4).

The average pH of sample sets from all producers exhibited the expected pattern―the pH of concentrate was slightly higher, and permeate slightly lower, than that of the starting sap material (Table 2). There was no trend for the pH of concentrate with increasing concentration level. However, as would be expected, the concentration of all constituents increased with increasing concentration (Table 2). Reflecting that, the conductivity of concentrate samples increased with increasing concentration levels. There were no trends in the pH or conductivity of permeate samples with increasing concentration level (Table 2).

The carbohydrate and mineral compositions were analyzed for the permeate samples from 2 producers, C and L (Table 2). These data illustrate the presence of Ca, K, Na, and Mn, and highlight the fact that membranes will allow the passage of some mineral ions [12] . However, the retention rates for these minerals in these samples were still quite high, with averages of 99%, 96%, 61%, and 99%, respectively.

Table 3 shows the average composition of permeate, sap, and concentrate samples at early, middle, and late timepoints of the maple production season. The

values for sap samples exhibit some patterns across the season, and the data are also presented graphically to illustrate these (Figure 1). Sap sugar content as measured by a refractometer remained relatively stable across the production season (1.9, 2.2, and 1.9 ˚Brix, respectively). However, the carbohydrate data illustrate more nuanced changes, with sucrose concentrations rising slightly at mid-season before decreasing in late-season samples, and glucose and fructose concentrations increasing in late-season samples (Table 3, Figure 1). Conductivity generally increased from early to late season, and, likewise, the concentration of most minerals (Ca, K, Mg, Mn, and S) also generally increased. In contrast, concentrations of both P and Na decreased over the season, and those of

^†Four sap samples with outlier values for trace metals (Fe, Cu, Zn) were removed from mean calculations to prevent obscuring seasonal patterns.

trace minerals Fe, Cu, and Zn showed no discernible seasonal pattern (Table 3, Figure 1). The increase in Mn over the course of the season is interesting to note, as Mn has been putatively associated with early-season membrane fouling observed by some producers, particularly at higher concentration levels (Carl Lapierre, personal communication). However, that Mn concentrations in sap were actually lowest at the beginning of the season suggests that the relationship between Mn and this early-season membrane fouling may be indirect or noncausal. The properties of permeate samples were similar across all timepoints

Sucrose was not included in this analysis, as it is confounded with concentration factor; B and Na were also excluded, as the elements were below the detection limit in sap and concentrate samples, respectively.

(Table 3). Seasonal patterns cannot be examined for concentrate samples because the values are comprised of samples from such a wide range of concentrations.

Concern is sometimes expressed that RO could potentially allow the loss or disproportionate concentration of sap constituents, altering the fundamental composition of maple sap, and thereby resulting in unacceptable effects on the composition of syrup. However, if RO units and membranes are operating optimally, they should essentially simply concentrate sap, removing water without significant loss or addition of constituents present in the sap. If this is the case, the composition of constituents in concentrate should be very similar to what is predicted by the concentration factor between the initial sap material and the concentrate. The data in this study can be used to assess this, by comparing the actual concentration of constituents in the concentrate to that predicted by the concentration factor. Table 4 shows the average differences between the actual composition values of constituents in concentrate samples and the predicted concentration of each constituent based on the sap to concentrate concentration factor. These data illustrate that the differences between actual and predicted values are typically quite small, even for constituents known to pass through membranes in small amounts, such as K and Ca (Table 4). They also illustrate a substantial amount of variation, with all constituents being present in both higher and lower concentrations than predicted by the concentration factor in some samples, as illustrated by the minimum and maximum difference values (Table 4). Of particular interest are the results with respect to “invert sugar”, glucose and fructose. It is often presumed that concentrate will have a proportionally higher invert concentration than the raw sap from which it was generated due to increased microbial activity with greater sugar concentrations. However, the concentrate samples in this study contained similar quantities of glucose and fructose to those that were predicted by the concentration factor, with average differences reflecting lower concentrations in concentrate than predicted by the concentration factor (Table 4). This is consistent with results observed in a previous analysis of sap and concentrate composition from a single RO [8] , and suggests invert levels are not typically disproportionately elevated in concentrate. This analysis does not preclude the existence of effects on other constituents not analyzed, such as phenolics or amino acids; however the size and properties of most other sap constituents would make significant loss or disproportionate concentration unlikely. And although variability was observed, in general this analysis suggests that reverse osmosis in maple production functions generally as simple concentration, not altering the fundamental proportional composition of maple sap.

An additional analysis was used to determine if the level of concentration used was related to reverse osmosis performance. Statistical analyses were conducted to determine if any significant relationships existed between the concentration factor and the difference between the predicted and actual concentration of any constituents in the concentrate. For most constituents, there was no significant relationship between the concentration factor and the differences between the predicted and actual concentration of the concentrate (Table 4). However, the differences between the predicted and actual concentrations of Mn were significantly related to the concentration factor, with more negative values at higher concentration levels (Table 4). This indicates that at higher concentration levels, there tended to be more Mn in concentrate than predicted by the concentration factor. This could suggest that the interaction of Mn (or a Mn-containing sap constituent) and RO membranes might be altered in some way at higher sap concentration levels, leading to larger quantities than predicted by the concentration factor to be retained in the concentrate. If true, this could have important implications for RO operation at higher sap concentration levels, and the mechanism through which this occurs would be critical to elucidate. However, although this analysis is suggestive, it is possible that the relationship is confounded by some factor not accounted for, or by inherent errors or biases in the analyses. The behavior of Mn and interaction with RO membranes at higher concentration levels should be further investigated.