The three regions selected for the study were the Lower Fraser Valley (LFV), Vancouver Island and Coast (VI) and Okanagan (OK), in BC Canada (Figure 1). These areas represent the most productive agro-climatic zones, ranging from dry to wet regions within BC. The OK basin is one of the driest in southern Canada, with a semi-arid climate; the LFV is one of the wettest watersheds with a moderate, oceanic climate; and the VI has a temperate-Mediterranean climate (Figure 2) [31] [32] [33]. The three study areas are situated in different agro-climatic zones, as they differ in growing degree days, and their variability in precipitation during the growing season is directly related to their irrigation water demand (Table 1) [34].

*Okanagan growing season was considered from May to September. Vancouver Island and the Lower Fraser Valley from April to September.

The following crops were selected for the study; blueberry (Vacciniumcorymbosum L.), raspberry (Rubusidaeus L.), strawberry (Fragaria × ananassaDuch.), potato (Solanumtuberosum L.), sweet corn (Zeamays L.), pumpkin (Curcubutapepo L.), green peas (Pisumsativum L.), apple (Malus × domestica Borkh.) and grapes (Vitisvinifera L.), as they are representative of field crops, fruit trees and berries grown in the OK, VI and LFV. These crops represent a range of both annual and perennial crops that vary in management practices and physiology (e.g., berries, peas and sweet corn aboveground consumables and below ground edibles such as potato).

Nutrients selected included three that a deficiency may be alleviated by soil management, namely plant macronutrients nitrogen (N) (estimated through protein content), phosphorus (P) and potassium (K), and one that is manufactured within the crop or a value-added factor, vitamin C. Constituent water and caloric energy content in the crops were also assessed. Water demand for different irrigation techniques used in the three regions was calculated to show how varied irrigation management can affect water demand for each crop in each different agro-climatic zone.

For each region and crop, the water demand and nutrient content were calculated on an annual basis, as graphically described in Figure 3 and consistent with the procedures outlined by [30] and [37].

For the soil characteristics, the Soil Management Groups considered were those that could grow the crops selected for analysis in each region (Table S1) [38] [39] [40]. Two soil series with different textures were chosen for each group [41] [42] [43] [44]. The number of soil series selected was 18 for the OK, 12 for the LFV and 10 for the VI. For each soil series, one location/parcel identification (ID) given on the BC Soil Information Finder Tool (SIFT) [45] was selected to calculate the water demand (Figure 3).

For the specific calculations, the parcel IDs were used in the BC Agriculture Water Calculator v2.1.1 [46] as a comparative indicator of the virtual water for each crop, soil series (determined by the soil texture input in the model) and two irrigation systems—sprinkler and drip (Figure 3). Sprinkler irrigation refers to a system where water is sprayed into the air onto crops using pumps, hoses, and sprinklers, while drip irrigation refers to a system that slowly dispenses water from irrigation tubes with regular punctures over the crops rooting zone.

The total annual production (kg/yr) for each crop and region was calculated by multiplying yield values (kg/ha) by the harvested areas (ha). Annual yield (kg/ha) from averages of BC Fast Stats data between 2015 and 2019 [36] and harvested areas (ha) from the 2021 Agricultural Census were used in the calculations [47]. The LFV total area was based on the sum of Fraser Valley and Greater Vancouver districts, while the VI area included the total area for all Vancouver Island and Coast districts and the OK the total area for all Thompson-Okanagan districts [35].

To examine virtual water in relation to local yield (kg of water/ kg of crop), the virtual water results from the BC Agriculture Water Calculator (m³/ha) were divided by the yield (kg/ha) of each crop and multiplied by the density of water (999.07 kg/m³ at 15.6˚C [48]) (Figure 3).

The nutrients, constituent water and caloric energy contents for each crop were based on the Canadian Nutrient File (CNF) [49]. Annual nutrient contents were calculated by multiplying the content per kg of crop by the annual production (Figure 3).

Annual food wastage was calculated based on an estimate of 30% of total food loss and waste, which is a conservative estimate for Canada [10], and was used in the calculations of nutrient and virtual water losses (Figure 3). The annual wastage for the crops considered in the study were compared to daily nutritional guidelines: 2600 kcal/day for 30-year-old males [50]; 64 g/day of protein for 80 kg adults [51] and 90 mg/day of vitamin C [52]. The estimated virtual water losses were compared to the daily residential water use in BC [53]. Fertilizer estimates were based on nutrient content N, P and K in associated common fertilizers in BC. These include ammonium sulfate (20-0-0 24S) and urea (46-0-0) for N fertilizers, monoammonium phosphate (11-52-0) for P fertilizer and potash (0-0-60) for K fertilizer [54]. The total losses were divided by the respective annual nutritional requirements and annual water use to get the number of people per year that could be supplied by the food wastage in the regions, and by the total harvested area to determine the amount of fertilizer that was wasted on the land area.

To examine the driver factor of virtual water, a principal component analysis (PCA) was completed using the FactoMineR package [55] on soil properties using texture (% sand, silt and clay), organic carbon content (%), soil bulk density, growing season and virtual water for each crop, to assess any variability in these properties inherent in the different regions. Organic carbon, bulk density and soil texture for each soil series were gathered from the SIFT [45].

To examine the return per m³ of water used in the three regions, the total sales of each crop were divided by the estimated total virtual water. First, the sales rates per ton of crop were calculated by dividing the total farm gate in BC by the BC marketed production in tons [56]. Then, it was converted to sales per kg by dividing by 907.185 kg/ton [57]. The total sales value (M$/yr) in the three regions was calculated by multiplying the annual production of each crop by the sales per kg. Finally, the return per m³ of water was calculated by dividing the total sales value (M$/yr) by the total sprinkler virtual water (Mkg/yr) times 999.07 kg/m³ (based on [48]).

The LFV had the highest total production of the crops selected (167 million kg), with blueberry and potato showing the two highest production values. The OK was the second highest producer (141 million kg), with grapes and apples as the two highest production values. The VI had the lowest production of the crops selected (16 million kg), with potatoes and apples showing the two highest production values (Table 2). The crops selected represent 16% of the total harvested area in the three regions, as the analysis focused on crops for human consumption. Since the three regions show hay and forage as the majority of their agricultural production [35], the impact of food waste and water allocation are potentially much higher than estimated.

The virtual water estimates varied among the three regions. The PCA found that 88.1% of the variability in the data showed a response to the grouping of regions (Figure 4). The LFV and VI had similar growing season lengths, and their ellipses overlap, while the OK had the driest climate among the three regions and the highest virtual water (Figure 4). Additionally, comparing the virtual water under different soil textures, the OK soils had virtual water up to two times higher than the LFV and VI for the same soil texture (Figure 5). This suggests that climate parameters such as the growing season have a higher influence than soil texture on virtual water estimations.

The OK region had higher variability in soil texture, and the coarser soils were the ones that varied the most when compared to the LFV and VI virtual water estimates (Figure 5). BC already faces water deficits, especially during summer, and with increasing climate variability, the OK is likely to experience more water shortages [24] [25]. Consequently, selecting finer soils, especially for growing crops with relatively high virtual water, presents a more compatible strategy for

water scarcity scenarios in the OK region.

Sprinkler irrigation had an average annual virtual water 28% higher than drip for all crops and regions studied (Figure 6), which was expected as sprinkler systems lose more water due to wind and evaporation [58]. Most farms in BC currently use sprinkler irrigation, so transitioning to more efficient systems such as drip irrigation could significantly reduce local water use. This is an important opportunity, especially for drier areas such as the OK, since drip irrigation systems are suited for fruit crops and can be adapted to varied soil conditions [59] [60].

For all the crops selected, the OK had an average virtual water almost two times higher than the LFV and VI (Figure 6). This is consistent with the results from the PCA (Figure 4), with the LFV and VI showing similar groupings and the OK with a different trend. The differences in virtual water among the three regions can be explained by the hotter and drier climate in the OK. The higher evapotranspiration rates during the hot dry summer of the OK lead to higher crop water needs and higher total virtual water [21].

In the three regions studied, pumpkins and apples had the highest virtual water per hectare among the crops selected, while grapes and sweet corn had the lowest (Figure 6). However, the ranking of crops changed when the yield was considered (Figure 7). Pumpkins and apples had some of the lowest virtual water per kg of crop, while grapes and sweet corn virtual water were relatively higher (Figure 7). Detailed virtual water data are presented in Table S2 and Table S3. The differences in virtual water in relation to yield were consistent with global assessments, with potatoes and pumpkins showing lower virtual water than berries and peas [20]. This suggests that virtual water in relation to production

would be useful to consider in future studies on food loss and waste.

In the three regions, grapes had a virtual water divided by yield almost two times higher than apples (Figure 7). This can represent a potential issue for water allocation in the OK region since grape production has increased by 36% and apple production decreased by 10% from 2016 to 2021 [47] [61]. In addition, the grape virtual water was estimated without a cover crop. With cover crop, the water demand is approximately 45% higher for sprinkler systems in those regions, based on the BC Agriculture Water Calculator [46].

Blueberry showed a relatively high virtual water per hectare and per kg of crop produced (Figure 6 and Figure 7). The blueberry production in the LFV increased by 16% from 2016 to 2021, while hectares of other berries dropped (for example, raspberry by 36%) [47] [61]. These changes are related to profitability but may enhance the water demand in the region in the long term [62]. Thus, measuring virtual water is a useful concept for assessing water management by comparing different land uses on a watershed or river basin scale [63].

For the crops selected in the study, the LFV had the highest total virtual water, with 39.4 billion kg of water per year for sprinkler irrigation; the OK had the second-highest, with 35.4 billion kg of water per year for sprinkler irrigation; while the VI had a much lower total virtual water, with 2.2 billion kg of water per year for sprinkler irrigation (Table 3). However, the VI showed higher return per m³ of water ($81/m³) than LFV ($78/m³) and OK ($37/m³) when adding the return from the nine crops selected for the study (Table 4).

Comparing the LFV and OK, the OK had a lower total production (Table 2) and lower total sales (Table 4) but used relatively similar volumes of water compared to LFV to produce the nine crops selected. Blueberries and potatoes accounted for $228 million and around 34 Mm³ of water in the LFV, whereas grapes and apples accounted for $169 million and around 33 Mm³ of water in the OK (Table 4). The LFV had a total production for the crops selected 19% higher than OK (Table 2), but the total virtual water was only 11% higher (Table 4).

*Potato data obtained from Potato Market Information Review reports [64].

Consequently, the LFV had higher returns per m³ of water than the OK (Table 4), which means that the water used in the OK brought less revenue than in the LFV for the nine crops selected.

The nine crops selected also varied in return per m³ of water within each region. Blueberries had a lower return per m³ of water than the other berries studied and a higher virtual water per kg of crop (Table 4). If blueberry production continues to rise in the LFV, this may represent a risk for local water allocation for this region in the longer term. As for the OK, grapes were the more economic choice than apples, with higher revenue per ton and higher return on water invested (Table 4). However, the virtual water per kg of crop for grapes is almost two-times higher than apples (Figure 7). Thus, as this industry continues to grow, the pressure on summer water withdrawals may continue to rise in the OK region.

The different crops showed different water requirements and the different irrigation systems provided varying efficiencies (Figure 7). Therefore, agricultural management practices can influence water resource availability in the long term [65]. To mitigate and also to prevent water scarcity scenarios, selecting crops better aligned with the local seasonal water availability will be important for the three regions. Lower water demand crops, such as apples and potatoes, finer soils, and more efficient irrigation systems such as drip are more congruent with water stress scenarios. To make such decisions, the BC Agriculture Water Calculator is a useful tool to help land managers and policymakers to meet local water demands.

Food quality indicators differed among the crops selected and are presented on a weight basis, e.g. 0.76 kg of constituent water per kg of sweet corn (Figure 8). Constituent water had a lower variability, ranging from 0.76 kg/kg for sweet corn to 0.92 kg/kg for pumpkins. Caloric energy had the opposite trend, with pumpkins showing the lowest content of 260 kcal/kg and sweet corn the highest with 860 kcal/kg (Figure 8). This was expected since crops with higher water in their composition are less calorie-dense.

Protein contents varied significantly among the crops selected, with apples showing the lowest value with only 2.6 g/kg and green peas the highest with 54.2 g/kg. Annual vegetables such as green peas, sweet corn and potatoes had higher caloric energy and protein contents than the perennial berries studied (Figure 8). Additionally, protein may be used as a proxy for N content based on nitrogen-to-protein conversion factors for different food types, commonly ranging from 4.4 to 6.25 [66].

Vitamin C contents were similar for apples, grapes, pumpkins and blueberries, ranging between 0.03 g/kg for grapes and 0.10 g/kg for blueberries, while raspberries and potatoes were higher with 0.26 g/kg and 0.20 g/kg, respectively. Green peas had vitamin C content almost two times higher at 0.40 g/kg, and strawberries had the highest with 0.59 g/kg. P and K contents were higher for the vegetables than for the fruits studied. Green peas had the highest P content (1.08 g/kg) and potato the highest K content (4.21 g/kg) (Figure 8). Local productivity, climate conditions and soil management would also likely influence the estimates of energy, constituent water and nutrients for each crop; however, additional primary nutritional data for each crop and within each region would be needed.

Around 30% of the food production in Canada is wasted throughout the FSC [10]. Considering the nine crops selected, almost 100 million kg of food is not supporting human nutrition in the three regions studied (Table 5). In Canada alone, 35.54 million metric tons of food is wasted every year [28] and the retail and final consumers account for more than half of that (20% out of 30%) [10], mainly associated with improper storage, excess purchases, misunderstanding of expiration dates and high aesthetically-pleasing food expectations [67]. To address this, Canada committed to the United Nations 2030 Agenda of Sustainable Development in 2015, supporting the target to halve the global food waste per capita at the retail and final consumer levels [68].

Different compositions of each crop make their impact vary among the food quality indicators selected for the study (Figure 9). Due to their composition, sweet corn and green peas showed higher losses of protein and P, while strawberries had higher losses of vitamin C in the three regions. Potato accounted for a high share of the waste and loss of all the indicators in the three regions (Figure 9). In

the LFV, blueberry had the highest loss of constituent water and caloric energy, while potato had the highest total loss of protein, vitamin C, P and K. In the OK, apples had the highest loss of all indicators except for protein which had grapes as the highest. In the VI, potato had the highest total loss of all food quality indicators studied (Figure 9).

Total loss and waste of the crops selected in the three regions would be enough to meet the daily guidelines requirements of caloric energy of more than 60,000 adults per year [50]; the protein requirements of more than 40,000 adults per year [51]; and the vitamin C requirements of more than 300,000 adults per year [52] (Figure 9). The estimated total virtual water wastage (30%) in the three regions added up to 23 billion kg of water for sprinkler irrigation and 18 billion

kg of water for drip (Table 3). This estimated virtual water loss corresponds to the annual residential water use of more than 180,000 people in BC [53]. This confirms the significance of food wastage impacts on local food security and local water demand within a regional perspective.

Macronutrients, including N, P and K, are needed by plants in relatively large quantities and are typically added annually to replace nutrients lost to crop harvest. The total P loss and waste in the three regions accounted for 26 × 10³ kg, corresponding to 50 metric tons of monoammonium phosphate, while K loss and waste added to 190 × 10³ kg, corresponding to 317 metric tons of potash (Figure 9). The P and K loss and waste would be equivalent to applying monoammonium phosphate and potash fertilizers at a rate of 2 and 13 kg/ha respectively for the total area studied. The total N may be estimated as 16% of the protein contents, accounting for 154 metric tons of N wasted [66]. This corresponds to 768 metric tons of ammonium sulfate or 334 tons of urea and would be equivalent to applying ammonium sulfate or urea fertilizers at a rate of 32 or 14 kg/ha respectively for the total area studied. The loss and waste of these nutrients have important implications for soil management and fertilizer demand since chemical fertilizers are resource-intensive to produce and are becoming limited worldwide [69] [70].