Turn the Tap: Integrated Research to Support Sustainable Irrigation Practices on Northeast Vegetable Farms

Our results are drawn from data collected in VT in 2019/2021, and ME in 2021/2022. Analysis is being finalized as of this final report, though many of our findings are presented below. All data will be included in a manuscript submitted for scholarly review and publication in winter 2022/spring 2023. Below is a summary of our analysis to date.

Soil parameters: 

Water use: In Vermont and Maine, the amount of water used in each treatment was significantly different from the control treatment (R2 =0.540, P < 0.001), as was expected. Timer treatments received significantly more water than the control treatment (R2 = 0.61, P <0.001), as did the feel of soil treatment (R2 = 0.61, P = 0.004). However, soil moisture sensor treatments did not significantly differ from the control treatment (R2 = 0.61, P=0.052). In Vermont, the soil moisture sensor had an estimated coefficient of 68.6 gallons of water where the timer treatment and feel of soil treatment has an estimated coefficient of 787.9 and 102.9 respectively. A Tukey post hoc test established that the timer treatment used statistically more water than all other treatments, while the feel of soil and soil moisture sensor did not statistically differ from one another. Analysis of the Maine data is ongoing.

N in leachate: Vermont leachate data showed that the timer treatment (R2 = 0.32, P< 0.001) and the feel of soil treatment (R2 = 0.32, P=0.0453) were different from plots in the control treatment. The timer treatment plots leached 74% more water than the control treatment plots, and 90-94% more water than feel of soil or soil moisture sensors plots. There was no leachate collected from the control treatment in Vermont in 2019, and all leachate collected from the control treatment in 2021 came from one plot (indicative of a leak in the irrigation system). In Maine, there was no difference in the amount of leachate collected between treatments. There was leachate collected from all plots, with the largest amount from the timer treatment, with an average of 3.936 L per plot over the 2021 growing season. Meanwhile, the control treatment plots averaged 1.333 L. There were several periods in 2021 in Maine in which the trial was saturated, with standing water. The soil texture caused slow drainage, and during this period leachate collected was similar across treatments.

In Vermont, irrigation had a large effect on nitrogen-in-leachate. The timer treatment plots had significantly greater concentration of nitrate in the leachate and significantly fewer nitrates in the soil than those samples taken from control plots (R2 = 0.034, P = 0.0163). The amount of irrigation water that drained through the soil both increased the amount of water leaving the root zone and transported N. Soil moisture sensor plots had significantly fewer soil nitrates in leachate than control plots (R2 = 0.034, P = 0.039) but did not differ in the amount of leachate collected.   

Nitrogen lost to leachate was calculated: total nitrates lost per lysimeter area (NL) was determined by multiplying the volume of leachate (L) and the concentration of nitrate in the leachate (NinL). Once total nitrogen loss (NL) was calculated, this value was multiplied by the square feet of a plot (Psqft) then divided by the square feed of the lysimeter (Lsqft). This resulted in nitrogen lost in whole plot (NP). The NP was then multiplied by 100 and divided by the amount of nitrogen to each plot (F), equaling the total N lost through leaching (LostN). A GLM established differences between each treatment compared to the control. In Maine, no treatment lost statistically more N to leachate than the control. In Vermont, all plots with irrigation lost more N to leachate compared to control. However, it should be noted that in 2019 there was no leachate collected from control plots and therefore we could not conduct a comparison to other treatments.

Yield:

Cucumbers: Cucumber yields in Vermont and Maine were statistically different in 2021 (R2 = 0.772, P < 0.0001), therefore the results for the two sites are presented separately. In Maine, there were no statistically significant differences in cucumber yield between any of the treatments. The control treatment had a mean of 56 individual cucumbers (36lbs), averaged within treatments for the growing season; the timer treatments yielded a mean of 55 cucumbers (34lbs). The feel of soil and soil moisture sensor treatments yielded a mean of 61 (38lbs) and 63 (39lbs) cucumbers respectively. This indicates that in Maine in 2021, none of the irrigated treatments outperformed the control treatment. In Vermont across both years, there was also no statistically significant difference in number of fruit when comparing irrigated treatments to the control treatment. However, control, timer and soil moisture sensor treatments had the low number of fruits with a mean of 91 (64lbs), 94 (68lbs), and 85 (62lbs) cucumbers averaged within treatment, while the feel of soil plots yielded a mean of 133 cucumbers (124lbs). 

In Vermont in 2019, the feel of soil treatment yielded more cucumbers than the control treatment (R2 = 0.55, P = 0.036). This could be due to 20 cB being a suboptimal soil dryness threshold for cucumbers, or that the soil moisture sensors were not at an optimal depth for assessing soil moisture for cucumbers. Cucumbers have relatively shallow roots compared to tomatoes and peppers, and sensor placement was likely lower than cucumber roots reached. In 2021, there was no difference in yield between any of the treatments. The likely difference in trends between the two years was the seasonal distribution of precipitation. In 2019, the majority of precipitation occurred in spring and early summer, prior to the growing season. 

It is worth noting that total precipitation across the 2021 growing season was relatively evenly distributed in Maine, which likely diminished the effects that different irrigation treatments had on cucumber yield at this site. Additionally, there are contradictory findings presented in prior studies on the effects of irrigation in cucumbers. Some suggest that key physiological indicators of cucumber development (e.g., leaf elongation and stomatal conductance) are less affected by soil water depletion than one may expect. In other words, it would take an extended, or several successive dry periods before physiological harm would be observed. In the case of the research presented in this present study, it is also possible that cucumber water uptake, and by extension plant fitness and yield, was affected more by environmental factors such as temperature or disease rather than irrigation treatment.

Peppers: Of the three crops included in this trial, peppers had the most variable yield across all treatments in both sites and both years. A GLM with gaussian distribution showed that Maine and Vermont sites in 2021 were not statistically different from each other (R2 = 0.477, P = 0.361); therefore, data was pooled for analysis. Both years in Vermont were also statistically similar, therefore yield data was again pooled (R2 = 0.293, P = 0.318). All years and sites were also analyzed separately.

Pooled data from Maine and Vermont in 2021 indicated all irrigated treatments yielded less than the control treatments. The largest difference was between the control treatment and the timer treatment (R2 = 0.565, P = 0.021). Over the growing season, the control treatment yielded a mean of 41 peppers per plot, averaging 12.6 lbs. The feel of soil, soil moisture sensor, and timer treatments yielded 33 (9.5lbs), 34 (9.7lbs), and 25 (6.7lbs) peppers respectively. The timer treatment was the only treatment that yielded significantly fewer peppers than the control plots (R2 = 0.246, P = 0.019). These results may suggest that a pepper yield is affected by overwatering. Previous irrigation studies support the idea that peppers are sensitive to soil moisture status. In Vermont there was sufficient water through precipitation for pepper growth, adding additional water through soil moisture sensors or feel of soil treatment did not increase yield and adding water daily decreased the yield of peppers in 2021.

In 2021 trial in Vermont, the feel of soil and timer treatments yielded significantly fewer individual fruits and lower total weight than the control treatment. The feel of soil treatment yielded a mean of 26 (8.1 lbs) peppers per plot over the growing season, the timer trial yielded 24 (6.4 lbs), and the control treatment yield double at a mean of 52 (16.2 lbs) peppers over the growing season.

Tomatoes: Tomato yield differed between Maine and Vermont in 2021 (R2 =0.434, P= 0.017) so data was not pooled for analysis. Vermont yield data in 2019 and 2021 did not differ significantly (R2 = 0.316, P = 0.360) and was pooled together, though each year was also analyzed separately.
In Maine, there was no difference in the total number of fruits averaged per plot over the growing season between any of the irrigation treatments and the control (176, 58 lbs). In Vermont, when data was pooled across 2019 and 2021, the timer treatment yielded statistically fewer tomatoes by fruit count per plot than the control plots (R2 = 0.194, P= 0.045). In Vermont the control treatment yielded 200 tomatoes per plot over each growing season, but with a smaller tomato size on average. Though the control plots had 21 more tomatoes on average, the mean weight was only 35.7lbs. This is consistent with prior research showing that timer systems can lead to overwatering causing leaching of inorganic nitrogen, thereby decreasing yield relative to more water-efficient irrigation practices. Both the feel of soil and soil moisture sensor treatments had comparable fruit counts with no statistical difference between them (189 and 185 fruited on average, respectively) with average total fruit weights of 40.7lbs and 37.3lbs respectively. When Vermont 2019 and 2021 were analyzed independently of one another, there were no significant treatment effects on tomato yield.

Quality: Crop quality (measured by assessments of color uniformity, gloss, shape, firmness, and number of defects) varied across sites and years.

Cucumbers: Though treatments did not have an observable effect on cucumber yield, there were significant differences in cucumber quality in Maine. Fruit harvested from plots with no irrigation (i.e., control treatments) were more uniform in color and had fewer defects but were significantly less firm and less uniform in shape and size compared to the control treatment. In examining the quality of cucumbers in Vermont, the gloss was the only quality parameter affected by irrigation treatments. The soil moisture sensor treatment was significantly less glossy than the control treatment, which received no irrigation (R2 = 0.05, t = 2.524, P = 0.036), however the difference was small (𝛽 = 0.052). With the contradictory quality results, the effect of irrigation strategies on the quality of cucumber in Maine in 2021 is unclear. In the context of the literature, more water is not necessarily better for cucumber quality. Indeed, some studies show that quality parameters (i.e., sugars, vitamin C) decline as water applications increase. This is a finding that is supported in research in Mediterranean climates, which shows that reduced water applications lead to a higher concentration of sugars and higher acidity, which are thought to contribute to desirable flavor profiles. These findings suggest that aligning the quality parameters used in the present research (developed by the USDA AMS) with parameters used in other crop research studies (i.e., sugars, acid, and water content) could lead to more understandable (and useful) results. 

Peppers: The quality of peppers varied very little over both year and location, however in Vermont in 2021, the feel of soil treatment resulted in less gloss than the control treatment. These results show that in the Northeast our changing and unpredictable precipitation patterns can have a mixed effect on the quality of peppers from year to year, even if yield effects were not particularly noticeable over the two years in Vermont and two locations. This finding aligns with prior research that shows that irrigation water quantity has a significant effect on pepper yield, though other studies have documented that deficit irrigation reduces yield compared to more generous irrigation approaches. To better tease apart these contradictions, future research should be conducted in a protected environment (e.g., greenhouse or high tunnel) which would better control external factors and make it more likely that differences between irrigation practices would become more observable.

Tomatoes: When comparing tomato quality parameters, there was no treatment effect in Maine in 2021. Though irrigation treatments did not have a statistically significant effect on yield in Vermont (in pooled data), there were significant differences in quality observed in 2019. Tomatoes harvested from the control plots exhibited lower quality than all other treatments which was not observed in Vermont 2021. These results suggest that even if irrigation does not increase yield of tomatoes, it may play a role in quality. This contradicts prior research suggesting that generally there is little difference in quality between tomato plants grown under drought conditions and non-drought conditions.

Studies that show minimal quality differences between different irrigation treatments. This is the basis of and justification for deficit irrigation, an approach designed to improve water use efficiency by limiting water availability to crops while maintaining an acceptable crop quality. Deficit irrigation is an evolving area of investigation, mostly in water-limited agricultural regions. In tomatoes, this approach is better suited to specific cultivars not included in this study, suggesting that the results presented here are likely influenced by the variety of tomato used. Future research should examine the relative fitness of different tomato varieties under the irrigation treatments included in this study, as it would provide useful information to Northeast growers about which varieties can tolerate water stress and/or water excesses. Because managing irrigation is a high-labor activity, it is often neglected on Northeast farms. Planting tolerant crop varieties may moderate the negative effects of suboptimal irrigation management.

Focus groups - qualitative findings: Farmers who attended the focus groups had a range of prior experiences with soil moisture sensors, from a general familiarity to no prior experience. Common themes that farmers discussed were (a) surprise to find they had been overwatering some crops, which was more commonly true in high tunnels; (b) preferences for various data delivery systems (in the field vs. in the cloud); (c) frequency with which they wanted to view soil water tension data; and (d) using soil moisture sensors to better manage "irrigation zones".

Overwatering in high tunnels was cited as a concern by some farmers. For example, one farmer who used sensors in a high tunnel where she grew tomatoes indicated that she was surprised to find out how much water she was actually using. This farmer paired sensors with a Netafim flow meter on her high tunnel irrigation line so she could assess how much water she was using to achieve desired cB readings. Blossom end rot was a challenge for this grower, and she concluded that overwatering was contributing to the problem. In the future, this grower planned to further fine tune her irrigation strategy to better control for consistent soil water content. The hope is that improved consistency would reduce tomato defects and improve the amount of marketable fruit produced in the high tunnel.

The desire to fine tune water applications was also mentioned by other focus group participants, as was the desire to "stop guessing" about how wet different soil textures were. This was discussed in the context of the cost of pumping water from wells, and the cost of purchasing municipal water. As one farmer stated: "I mean, it's money. We get double the yield with water, but we just don't know exactly how it went." Farmers in the focus groups voiced a desire to reduce unnecessary costs associated with accessing water and were interested in soil moisture sensors as a way to help them become more water efficient. They also noted that avoiding over-irrigation would help them make the most of expensive nutrient applications: "When you put your fertilizer in, how long do you water so you don't drive it past the roots?" is an example of a question that one strawberry grower had when thinking about assessing soil moisture during irrigation periods. He added: "We need to know much better what we're doing."

Data delivery systems for soil moisture sensors include integrated tension gauges, handheld meters to take readings in the field at a single point in time, and dataloggers that record continuous data. Data loggers can be connected to cellular or wireless networks, allowing farmers to access data from a smart phone or computer. Through focus group discussions, it became clear that farmers have different preferences for how to access data, with some preferring to read cB conditions in the field using a gauge or handheld meter. As one grower stated: "I don't hate the idea of a reader on the (sensor), because I'd have to run back to the house to get on the computer and login, because I couldn't do it through a phone. So, (a reader in the field would be) a quicker decision-making tool." Others indicated a preference for getting the information online. There is significant ongoing expense associated with cloud-based data, which some farmers indicated would be a barrier to them.  One farmer noted: "I'm okay with the prices (for the sensors) as they are. It's just that, if it were easier to get the data ... it depends on if you're going to get the software or not, honestly. That's the expensive part." Another stated: "I think the big question is: When does this pay? What carrot yield bump would justify this expense? You're getting fine carrots irrigation by feel(ing the soil) but would you get a $1,000 more of your carrot crop by using this technology? That's the think I keep coming back to." Other farmers indicated interest in alerts delivered directly to their phone, for both when to start and when to stop watering. 

Farmers in the focus groups indicated that they only wanted to check soil moisture sensors when they were able to irrigate. For some this was once a week, while for others this was every two or three days. Others indicated that their willingness to check sensors would be driven by weather. For high tunnel production, the farmer who grew tomatoes (above) reported that checking soil moisture levels was a daily ritual, but that was not necessarily true for sensors that she had installed in outdoor crops. Others indicated that having data that was collected on a shorter time step would be very useful to them. For example, one farmer stated: "I'm always curious about... water layering... turn it on for a little bit of time in the morning, and then once that's sunk down, then turn it on again... in order to get the moisture lower. I've always been curious about what's happening if I just water for ten minutes or what's happening if I water for an hour."

Farmers talked about using soil moisture sensors to identify irrigation zones. These participants indicated that parts of a single field or a single high tunnel could have very different moisture retention dynamics. In other words, some soils stayed wet while others dried out even in areas that were close together. One farmer noted that part of his field was located close to a wetland, and the water table was higher than in other parts of the same field. He stated: "Does it even matter if it's a little more saturated down there and a little bit drier up here if we're in a range? How differently does your soil have to respond before you reconsider your irrigation zones?"

Focus groups - survey findings:

Participants in the farmer focus groups were asked about how they assessed their irrigation needs in a typical year. Most (n = 9, or 69.23% of the 13 participants who answered the survey) reported that they irrigated based on crop condition and the feel of the soil. This aligns with prior surveys by our research team, and past USDA-ERS Irrigation Surveys. Additionally, five participants reported using weather reports (like frost warnings or heat advisories), and two reported that they did not irrigate. One person reported using a timer system on a valve. Most respondents (9 out of 13) had never used soil moisture sensors. Three participants had used soil moisture sensors in the past but had discontinued use because the sensors were "too complicated", "were defunct", or there was too much of a cost with "reinvesting on a new farm".

 Farmer participants were asked to report their level of agreement with several statements associated with on-farm water management before the workshop/focus group, and again after the workshop was complete. The workshop did not have an effect on participants level of concern about several measures (see table 3), but participants reported that the workshop increased their familiarity with soil moisture hardware and software, increase the likelihood that they perceived soil moisture sensor information as relevant to their farm operations, and their willingness to invest in soil moisture technology. 

We asked participants how much they would be willing to invest in soil moisture monitoring, if they could achieve 10%, 20%, 30%, or 40% increased yield. Several farmers reported that they would be willing to invest less than $500 for only a 10% increase in yield (8 out of 13 respondents), but willingness to invest greater sums was contingent upon higher rates of yield improvement. For example, 5 respondents would consider investing more than $2,000 if the return was a 40% increase in yield, but only 2 farmers would consider investing that sum for a 30% increase in yield. Only one respondent indicated that they would spend over $2,000 for a yield improvement of 20%, and no one was willing to invest this much money for a 10% yield improvement.

As a result of participating in the workshops and focus groups, farmers reported that their knowledge about soil moisture sensors had increased (4.5 average level of agreement, with 5 = strong agreement and 1 = strong disagreement). Participants also indicated that they would use soil moisture sensors in the future (mean response = 4.5) and that their confidence in using soil moisture sensors had increased (mean response = 4.0).

In summary, focus group findings showed that farmers are interested in soil moisture sensors as a way to improve water-use and nutrient efficiency, reduce unnecessarily irrigation, set up irrigation zones, and schedule irrigation cycles. These is a diversity of preferences among farmers when it comes to when they would like to access soil moisture data, and their preferred method of viewing the information. Cost of cloud-based data collection and storage is a barrier for some farmers, as the economic benefits of investing in these platforms has not been proven.