Final report for LNE21-433R
Project Information
This project systematically evaluated the effects of Late Water (LW) flooding on cranberry yield, fruit quality, and plant physiological traits over three consecutive years (2021–2023) in Massachusetts commercial cranberry fields (bogs) planted with 'Stevens' and 'Mullica Queen' cultivars. Each year, paired plots (LW vs. Control where no spring flooding was implemented) were established at grower-managed sites. LW bogs were flooded for ~30 days in spring, while controls were not.
Key findings:
-
Bloom Phenology: LW delayed bloom initiation by 10–13 days compared to controls across years and cultivars.
-
Fertilizer and Fungicide Use: Growers applied similar nitrogen and potassium inputs in late water and control plots, but applied ~18% less phosphorus and fewer fruit rot fungicide applications than controls.
-
Plant Physiology: LW reduced total non-structural carbohydrates (TNSC) during early growth, particularly in 'Mullica Queen'.
-
Canopy and Upright Metrics: 'Stevens' under LW had higher leaf area index (LAI), but LW plots had lower reproductive to vegetative upright ratios and number of berries overall.
Treatment Year Results:
-
LW significantly increased fruit rot incidence (~13% higher than controls) as against our hypothesis.
-
LW significantly reduced yield, especially in 'Mullica Queen', but differences in 'Stevens' were not significant.
-
LW increased berry anthocyanin content (TAcy) and slightly reduced fruit firmness.
Lag Year (Year after Late Water Flood) Results:
-
No significant difference in fruit rot incidence between former LW and control plots.
-
Yield increased in 'Mullica Queen' plots with LW history (~9,478 kg/ha higher), but no significant difference in 'Stevens'.
-
No significant effect on TAcy.
-
Fruit firmness was significantly lower in plots with LW history.
Overall, this work provides robust multi-year data on immediate and lag effects of LW on cranberry productivity and quality, supporting refinement of grower recommendations for sustainable disease management and quality optimization.
To evaluate the relative importance of critical criteria (e.g., field characteristics, water quality, crop status, pest and yield history, and environmental conditions) that contribute to the short- and long-term outcome of late water (LW) floods and to develop a web-based, data-driven, decision-making model (DMM) that will generate grower-friendly outputs that promote appropriate LW use. A LW-DMM would improve cranberry grower confidence and facilitate increased adoption of this under-utilized practice. Farm profitability will be enhanced through sustainable cranberry production, increased revenue, and reduced pesticide applications. This could positively impact >420 farms in the Northeast (NE), affecting >16,000 acres of cranberries.
Massachusetts holds significant status as the second-largest producer of cranberries (Vaccinium macrocarpon) in the United States, with approximately 14,000 acres under cultivation. Cranberry farms in Massachusetts are typically small-scale operations, with over 70% managing fewer than 20 acres. A persistent and significant challenge faced by cranberry producers in the region is the management of fruit rot diseases. The cranberry fruit rot complex, caused by more than a dozen taxonomically diverse fungi (Damm et al., 2012; McManus et al., 2003; Oudemans et al., 1998; Polashock et al., 2009; Weir et al., 2012), can result in substantial yield losses ranging from 50–100% if not effectively controlled. Economic penalties are often imposed when fruit rot exceeds 12%, with complete rejection occurring when rot surpasses 20%.
Flooding has traditionally been an integral part of cranberry production, used for winter protection against desiccation and frost damage. Observations of beneficial effects from occasionally extending winter floods led to the practice of late water (LW) (Franklin, 1948; Beaton, 1953). In modern cultivation, LW involves withdrawing the winter flood by March and re-flooding in April for 30 days (mid-April to mid-May) while vines are dormant. Initially, LW was considered especially beneficial for weak or old bogs (>30 years old with consistent pest pressure) (Beckwith, 1940). Historically, LW was crucial for managing pests and reducing fruit rot before widespread chemical fungicide use (Shear et al., 1931; Averill et al., 1997; Beaton, 1953).
LW benefits have been documented in native cultivars, notably 'Early Black' and 'Howes', including reductions in pests such as mites and fruitworms, suppression of weeds, synchronization of bloom, enhanced fruit size, and overall improvement in fruit quality (Averill et al., 1997; Botelho and Vanden Heuvel, 2006). Studies also indicate that LW-treated vines require fewer chemical inputs and less fertilizer. Despite these advantages, the adoption of high-yielding but rot-susceptible hybrids like 'Stevens' and the introduction of effective broad-spectrum fungicides (e.g., chlorothalonils and mancozebs) reduced reliance on LW. Additionally, anecdotal reports and limited empirical evidence suggesting occasional yield reductions with LW discouraged widespread use.
Recent shifts in pesticide regulations, particularly restrictions by the European Union (EU)—the largest importer of US cranberry products—have limited effective chemical options. In 2021, the EU lowered the Maximum Residue Limit (MRL) for chlorothalonil from 5 ppm to 0.01 ppm for fruit, and announced the non-renewal of mancozeb beginning in 2021, with a risk-based MRL review pending. These changes have forced growers to rely heavily on the few remaining effective fungicides (Caruso, 2011, 2012; Oudemans & Vaiciunas, 2009; Wells et al., 2014) that have single-site modes of action (FRAC Groups 3 and 11), increasing the risk of fungicide resistance (Giorgio et al., 2020). Combined with historically low prices and stringent quality demands, these factors have reignited grower interest in non-chemical strategies like LW.
Despite renewed interest, LW adoption remains limited due to uncertainty about its effects on newer cultivars and inconsistent anecdotal reports of yield losses during the LW year. Past research mainly focused on native cultivars and was often limited to single-season studies. Moreover, few studies assessed LW’s impact on key economic traits like yield, fruit rot incidence, total anthocyanin content (TAcy), and fruit firmness over multiple years, or considered potential lag effects in subsequent seasons.
Thus, critical gaps remain regarding LW’s consistent effects on contemporary cultivars like 'Stevens' (which accounts for over 40% of Massachusetts acreage), as well as on plant physiology, including TNSC reserves, nutrient dynamics, and interactions with environmental factors. This study aimed to systematically evaluate immediate and lag effects of LW on yield, fruit rot incidence, TAcy, fruit firmness, and other key traits in commercial production. For this, we studied 15 paired sites (late water vs. control) over three growing seasons (2021–2023), and assessed year-after effects in the same bogs (2022–2024), gathering data on fruit quality, weather, and cultural practices. We analyzed all data with the goal of understanding the impact of late water on fruit rot, yield, and quality in newer hybrids and developing a late water decision tool that provides data-driven, evidence-based guidance to help growers make informed decisions.
Cooperators
- (Educator and Researcher)
- (Educator and Researcher)
- (Educator and Researcher)
- (Educator and Researcher)
- (Educator and Researcher)
- (Educator and Researcher)
- (Educator and Researcher)
Research
Materials and Methods
2.1 Study Sites and Experimental Design
Field trials were conducted over three consecutive growing seasons (2021–2023) at commercial cranberry bogs (fields) located in southeastern Massachusetts. Each experimental year consisted of completely new sets of paired bogs: one bog per pair received the Late Water (LW) flooding treatment, while the corresponding paired bog served as an unflooded control.
In 2021, the experiment included six pairs of bogs (12 bogs in total), with all bogs planted with the 'Stevens' cultivar. In 2022, a different set of six paired bogs was studied (12 bogs total), comprising three pairs planted with 'Stevens' and three pairs with 'Mullica Queen'. In 2023, a completely new set of four pairs (eight bogs in total) was established, including two pairs planted with 'Stevens' and two pairs with 'Mullica Queen'.
Each paired site was labeled as Late Water (LW) and Unflooded Control (CON):
- 2021-LW-1 / 2021-CON-1: Plymouth
- 2021-LW-2 / 2021-CON-2: Carver
- 2021-LW-4 / 2021-CON-4: Wareham
- 2021-LW-5 / 2021-CON-5: Wareham
- 2021-LW-6 / 2021-CON-6: LW in Carver, CON in Middleboro (4 miles apart)
- 2022-LW-1 / 2022-CON-1: Wareham
- 2022-LW-2 / 2022-CON-2: Carver
- 2022-LW-3 / 2022-CON-3: Carver
- 2022-LW-4 / 2022-CON-4: Plymouth
- 2022-LW-5 / 2022-CON-5: Lakeville
- 2022-LW-6 / 2022-CON-6: LW in Wareham, CON in Carver (5 miles apart)
- 2023-LW-1 / 2023-CON-1: Wareham
- 2023-LW-2 / 2023-CON-2: Plymouth
- 2023-LW-3 / 2023-CON-3: Marion
- 2023-LW-4 / 2023-CON-4: Wareham
- 2023-LW-5 / 2023-CON-5: Carver
All paired sites were within 0.25 miles of each other unless specified above. Each paired plot consisted of similarly sized LW and control bogs, managed by growers following their typical practices except for the controlled flooding treatment. This experimental design allowed for robust comparisons between treatments, cultivars, and years, facilitating assessments of both immediate and subsequent (lag) effects of LW flooding.
2.2 Late Water Treatment, Cultural Practices and Environmental Data
Late Water Treatment: For the LW treatment, experimental bogs were flooded for approximately 30 days, typically from mid-April to mid-May, following standard cranberry management recommendations. Flooding was conducted using water sources commonly utilized by growers for harvest and winter flooding. Control bogs remained unflooded during this period.
Cultural Practices Data: Following water withdrawal, growers managed each bog according to their standard post-flooding and non-flooding practices.
At the end of each season, growers reported their management practices and inputs from each bog (both LW treatment and control). Fertilizer usage (N, P, and K) and fungicide applications (number of sprays) were particularly of interest. At the end of the season, total fertilizer rates and total number of fungicide sprays were calculated for each bog for each study-year.
Environmental Data
For each Study Year, we obtained weather data from nearby NOAA Weather Stations. Specifically, daily temperature, total precipitation (mm) and relative humidity (RH, %) were obtained.
2.3. Changes in Plant Non-Structural Carbohydrates and Phosphorus Dynamics
2.3.1 Total Non-Structural Carbohydrate Analysis
To assess vine carbohydrate status, cranberry uprights (current-season vine tissue) were sampled at two time points: just before LW flooding (early April, pre-flood) and immediately after flood removal (mid-May, post-flood). At each sampling, multiple handfuls of uprights were clipped randomly throughout each plot (avoiding edges) and combined into a composite sample per plot. Samples were oven-dried at 60 °C for 72 h and then ground to a fine powder using a Wiley mill. The ground tissue was stored in airtight containers at room temperature until laboratory analysis of non-structural carbohydrates.
Soluble sugars and starch: We quantified total soluble sugars and starch in the vine tissue using colorimetric assays following standard protocols. Approximately 50–100 mg of dried, ground tissue was extracted in boiling water and diluted acid, then filtered to obtain a clear extract for soluble sugars. Total soluble sugar content was determined with the anthrone reagent method, which reacts with hexose sugars to yield a blue-green color (measured at ~620 nm). Reducing sugars (e.g. glucose + fructose) were measured in the same extracts using the 3,5-dinitrosalicylic acid (DNS) assay, following Miller’s method (colorimetric reading at ~540 nm). Starch in the tissue residue was measured by acid hydrolysis (conversion to glucose) followed by the anthrone assay, thus quantifying starch as equivalent total glucose. Calibration curves were prepared with D-glucose for both anthrone and DNS assays, and results for sugars and starch were expressed as percent of dry weight. From these values, we calculated total nonstructural carbohydrates (TNSC). TNSC was defined as the sum of total soluble sugars and starch content. All assays were performed in triplicate per sample, and sugar/starch percentages were averaged for analysis.
2.3.2. Surface Water Monitoring and Phosphorus Analysis
To quantify phosphorus dynamics—specifically total phosphorus (TP) concentrations, fluxes, and exports—associated with late water flooding and standard grower practices, a study was conducted in 2022 in four of the experimental sites planted with MQ cultivar (three of which were flooded with late water and one serving as a control site). These sites were monitored for phosphorus loss. Sites wet harvested in the fall, with one site flooded twice in the fall, and 1-2 winter floods were applied to protect against winter desiccation. At each site, surface water depth was continuously (15-min) recorded using a pressure transducer corrected for barometric pressure (Model HOBO U20-001-04, Onset Computer Corporation) and referenced to field measurements of the depth of surface water relative to the soil surface (Fig. 1). Surface water flows were measured using acoustic Doppler velocimeters. Samples of surface water were collected in 500-mL polyethylene bottles that were prewashed in an acid bath (10% HCl) and rinsed with deionized water. Total phosphorus (TP) concentrations were digested with an alkaline persulfate reagent and analyzed at the University of Maryland Center for Environmental Science Nutrient Analytical Services Laboratory using EPA Method 365.1. TP fluxes were calculated as the product of surface water flows and total P concentrations in surface water. TP export was calculated as the difference in hydrologic outputs and inputs of TP (positive values indicate net loss of TP) for each flood.
2.4 Bloom Monitoring and Canopy Measurements
Bloom phenology and canopy development were monitored in each experimental bog (LW and Control) in all the study years to evaluate potential effects of LW on flowering and progression. Starting in early June through August, observations were made at approximately biweekly intervals. On each date, four random locations per bog were selected and at each location 10 uprights were examined. On these uprights, we recorded the number of flower buds, open flowers, pinheads, and developing fruit. These counts were used to calculate two bloom metrics of interest:
- Percent In Bloom = (Open Flowers) / (Flower Buds + Open Flowers + Pinheads + Fruits) × 100
Canopy leaf density was assessed at mid-season using a ceptometer to measure leaf area index (LAI). We used an AccuPAR LP-80 Ceptometer (METER Group, Pullman, WA) to take indirect LAI readings in each bog when the canopy was fully developed (late July to early August). On a uniformly overcast day (to ensure diffuse light), the ceptometer probe was inserted just under the vine canopy at multiple points per plot (5 readings evenly distributed), while a reference above-canopy reading captured incoming PAR. The instrument’s software calculated LAI from the light interception values, and an average LAI was obtained for each plot.
We also quantified the proportion of reproductive versus total uprights ((upright ratio = reproductive / (vegetative + reproductive uprights)) in late July. At that time, after fruit set, a 6-inch diameter circular frame (0.20 m² area) was placed at a randomly chosen fixed location in each plot. All uprights within the frame were clipped at the vine base and brought to the lab. Each sample was sorted into two categories: reproductive uprights (those bearing flowers or fruit) and vegetative uprights (those without evidence of flowering). The count of uprights in each category was recorded, and uprights ratio was calculated.
2.5 Harvest and Fruit Quality Assessments
At harvest time (late September to early October), fruit yield and quality were measured from each plot. Berries were hand-harvested from defined 1 ft² (0.093 m²) quadrats. Per bog, five 1-ft² sections were sampled randomly, and all berries within each quadrat were picked. The harvested berries were sorted and evaluated for yield components and fruit quality attributes as follows.
Yield and fruit rot: Berries from each plot’s quadrats were sorted to separate sound versus rotten. Percent rot was then calculated as number of rotten berries / total number of berries in square foot expressed as a percentage. Yield was determined by weighing sound berries per unit area and upscaling to kilograms per hectare (kg·ha−1).
Firmness: Fruit firmness was measured on a subset of sound berries using a FirmTech 2 Fruit Firmness Tester (BioWorks Inc., Wamego, KS). For each plot, ~25 representative ripe berries were selected randomly from the sound fruit pool for firmness testing, avoiding any damaged fruit.
Berry anthocyanin content: Total anthocyanin content (TAcy) was determined using a protocol from Fuleki and Francis (1968). Fruit juice was extracted by blending 100 g of fruit with 120 ml of 0.2 N HCl in a commercial blender for 10 s. Juice was extracted using Buchner funnel on a suction flask with a filter per. The filtrate was measured using a spectrophotometer at 515 nm and absorbance was obtained compared to the Aqueous color extraction test chart to determine the Tacy values.
Post Late Water Residual (Lag) Effects
To evaluate residual (lag) effects of LW treatment on the subsequent year’s crop, we continued observations in the year following each LW application. In practice, bogs that received LW in one spring were managed without LW in the next spring (since growers do not use LW on the same bog in consecutive years), but we still measured their performance in that next season under standard practices. The same paired bogs (former LW vs. former control) were thus studied for one additional year with no flooding treatment applied to either. During these follow-up seasons, we collected data for same fruit parameters of interest (yield, fruit rot incidence, fruit firmness, and berry anthocyanin content) as described above.
Statistical Analysis
Differences in Phosphorus export among floods were analyzed using a Kruskal-Wallis test, a non-parametric statistical test of significance at the 95% confidence level (p < 0.05) in JMP (v. 5) for Mac OS X Lion (v. 10.7). All other data were analyzed using the R statistical software environment (R v4.2.2; R Core Team 2023) (ref). Descriptive statistics (mean ± standard error) were generated for all measured parameters. A dependent two-sample T-test was performed on vine TNSC to test pre-flood sampling and post-flood sampling. While an independent two-sample t-test was used to compare management inputs (fertilizer and fungicide) between treatments and the changes in TNSC. uprights ratio, LAI.
The main treatment effects on plant and fruit variables were evaluated with linear mixed-effects models (LMMs). We fitted LMMs using the lme4 package in R, with treatment (LW vs. control) as a fixed effect.
Random effects were included to account for the blocked nature of the experiment: specifically, Year (2021, 2022, 2023) was treated as a random intercept to generalize conclusions across years, and Location was a random intercept to account for baseline differences among bog locations. Furthermore, Treatment interaction with fixed effects was initially included and assessed for their significance. Model selection was performed by stepwise elimination based on Akaike’s Information Criterion (AIC): non-significant terms and extraneous random effect terms were removed if this improved model parsimony without reducing goodness-of-fit (lowest AIC selected). Final models thus retained only significant parameters tested during model selection.
Model fitting was done by restricted maximum likelihood using lmer (from lme4), and lmerTest package was used to obtain p-values via Satterthwaite’s approximation for degrees of freedom. For each response variable (Percent rot, Yield, TAcy and Firmness), we report the LMM-estimated treatment means and the significance of the LW effect. We also assessed model fit by calculating the conditional R² for each mixed model, which represents the variance explained by both fixed and random effects. The conditional R² was derived following Nakagawa & Schielzeth’s method (2013) for mixed models, via the MuMIn package in R. All statistical tests were evaluated at α = 0.05. Data are presented with appropriate units, and all analyses were reviewed to ensure they met assumptions of normality and homoscedasticity.
Bayesian Network Learning
Bayesian networks were used to characterize potential causal relationships among field management practices, physiological variables, and the response outcomes of Percent Rot and Yield for cranberry production fields monitored from 2021 to 2023. To guide structure learning, variables were organized into logical tiers representing a hypothesized causal order, such that management factors and fixed characteristics were considered potential antecedents to physiological traits, which in turn could affect productivity measures. Based on agronomic knowledge, certain edges were explicitly disallowed through a blacklist to prevent biologically implausible connections.
The structure of the networks was learned using a score-based hill-climbing algorithm with the Bayesian Information Criterion (BIC) as the scoring metric. Network structure was estimated for the combined data set spanning all three years rather than fitted separately by year. To account for structural uncertainty, 200 bootstrap samples were generated and model averaging was performed across all bootstraps. During each bootstrap iteration, the hill-climbing algorithm added, removed, or reversed edges to maximize the BIC score within the constraints of the blacklist. Edges with an average strength greater than a pre-defined threshold (0.5) were retained in the final average network. The presence and direction of each edge were interpreted in terms of their relative frequency across bootstrap samples.
Once the final structure was established, network parameters were estimated as conditional Gaussian distributions for each continuous outcome given its parent variables. This approach produced local linear Gaussian models, analogous to multiple regression, that quantified the expected effect of each parent variable on the outcome. Conditional independence tests were also conducted to confirm that fitted parameters aligned with the learned network structure. Descriptive statistics, including means, standard deviations, and ranges, were calculated for all variables prior to model fitting.
All data preparation, structure learning, bootstrapping, parameter estimation, and visualization of the final directed acyclic graphs were conducted in R using appropriate packages for Bayesian network analysis.
3. Results
3.1. Experiment Sites Summary:
Over three years, the study examined 16 unique paired sites across southeastern Massachusetts. In 2021, six pairs (all 'Stevens') were tested. In 2022, six new pairs included both 'Stevens' and 'Mullica Queen'. In 2023, four pairs were studied (two 'Stevens' and two 'Mullica Queen'). Each pair had one LW treatment and one control site. Most sites were within 0.25 miles of each other, enabling robust comparison of immediate and lag effects of late water flooding on yield, rot, and quality metrics.
Bog Age
Bog age ranged from a minimum of 7 years to a maximum of 35 years across all study sites. The median bog age was 20 years, and the mean bog age was 22 years.
3.2. Cultural Practices Among the Experimental Sites:
Fertilizer Application
Total nitrogen (N) applied during the growing season did not differ significantly between Late Water and Control treatments (Figure 1); on average, approximately 2% less nitrogen was applied in Late Water beds compared to Control beds. Potassium (K) applications were also not significantly different between treatments, although Late Water beds received about 2% more potassium than Control beds overall. In contrast, phosphorus (P) application differed significantly, with Late Water beds receiving approximately 18% less phosphorus than Control beds. When examined by cultivar, ‘Mullica Queen’ bogs consistently received more fertilizer inputs for all three macronutrients (N, P, and K) compared to ‘Stevens’. Within cultivars, ‘Stevens’ beds under Late Water management had significantly lower phosphorus application than Control beds, whereas potassium application was significantly higher in Late Water ‘Stevens’ beds than in their Control counterparts.
Fungicide Application
Overall, a significantly greater number of fungicide applications were recorded in Control beds compared to Late Water beds during the treatment year (Figure 2). On average, Control beds received approximately four fungicide sprays per season, whereas Late Water beds received about three sprays. When broken down across all sites, 6 Control beds received three sprays, 8 received four sprays, and 2 received five sprays during the season. In comparison, for the Late Water treatment, 2 beds received two sprays, 8 received three sprays, and 6 received four sprays. Further analysis by year indicated that in 2021, 33% of paired beds (each Late Water beds with its paired Control) had the same number of fungicide applications, while in 2022 and 2023 this proportion increased to 50% of pairs having identical fungicide spray numbers.
3.3. Changes in Plant Non-Structural Carbohydrates and Phosphorus Dynamics
Total Non-Structural Carbohydrates (TNSC)
Both the absolute change in Total Non-Structural Carbohydrates (TNSC) and the percent change in TNSC were significantly different between Late Water and Control treatments (Figure 3-1 and 3-2). TNSC values showed considerable variation across individual bog pairs, but a consistent pattern was observed within treatments. Late Water beds showed a measurable decrease in TNSC during the treatment period, reflecting reduced carbohydrate reserves during early growth. In contrast, Control beds showed an overall increase in TNSC, as vines accumulated and restored carbohydrate levels following the standard winter flood used for frost protection. The magnitude of change in TNSC differed by cultivar. When results were analyzed separately, ‘Mullica Queen’ beds had a significant reduction in TNSC under Late Water compared to their Control pairs, while ‘Stevens’ beds did not show a significant difference in TNSC between Late Water and Control treatments. Across all sites, the direction of change in TNSC was consistent for Mullica Queen, while for Stevens the response varied more between individual bog pairs.
Phosphorus Dynamics
Results showed that the late water flood was a relatively minor source of P loss (Figure 3-2). On average, the late water flood resulted in P export of 0.03 kg P ha-1 yr-1, which was comparable to P export for the winter flood (0.01 kg P ha-1 yr-1), but 18 times lower than P export for the harvest flood (0.32 kg P ha-1 yr-1). These results are consistent with previous work showing that the harvest flood is a major source of P loss, especially compared with the winter flood (Kennedy et al. 2020). Collectively, these results point to seasonality as a control on P losses. The timing of the late water flood, which occurred early in growing season prior to any fertilizer applications, coupled with relatively mild air temperature of the spring, likely contributed to relatively low P losses.
3.4 Bloom Monitoring and Canopy Measurements
Bloom Progression and Initiation Dates
Bloom progression was monitored at intervals of 7 to 14 days to evaluate differences in floral phenology between Late Water and Control treatments, as daily visits were not feasible across all sites. For consistency, bloom initiation was defined as the date when each bed reached at least 5% open bloom. Across all years, bloom initiation was consistently delayed in beds that received the Late Water treatment compared to unflooded controls, indicating a phenological shift likely attributable to the extended spring flooding period. In 2021, ‘Stevens’ bogs under Late Water initiated bloom approximately 13 days later than their paired Controls. In both 2022 and 2023, a similar pattern was observed for ‘Stevens’, with a 12-day delay in bloom initiation between Late Water and Control. For ‘Mullica Queen’, the delay was slightly shorter, averaging 10 days. These results demonstrate that Late Water application consistently postpones bloom initiation across cultivars and seasons.
Upright Ratio
The upright ratio, defined as the proportion of reproductive uprights to total uprights on cranberry vines, was significantly different between Late Water and Control treatments (Figure 7). Across all sites and years, Control beds had a higher upright ratio compared to Late Water beds. This pattern was consistent when analyzed by cultivar. For both ‘Mullica Queen’ and ‘Stevens’, higher upright ratios were observed in Control beds than in Late Water beds, indicating a greater proportion of reproductive uprights under the Control treatment across cultivars and sites.
Leaf Area Index (LAI)
Leaf Area Index (LAI) did not differ significantly between Late Water and Control treatments when averaged across all study years (Figure 8). Across sites and years, mean LAI values showed similar canopy density in Late Water and Control beds overall. However, when analyzed by cultivar, different patterns emerged. For ‘Mullica Queen’, there was no significant difference in LAI between Late Water and Control beds, with values remaining comparable across paired bogs. For the ‘Stevens’ cultivar, LAI was significantly higher in Late Water beds than in Control beds, indicating a denser canopy under the Late Water treatment for this cultivar. Variability in LAI was observed among individual sites within each cultivar, but the overall trend for ‘Stevens’ showed consistently greater leaf area index under Late Water flooding.
Berry Number
Berry number per square foot showed the same pattern observed for the upright ratio across all study years and cultivars (Figure 9). The number of berries per square foot was significantly higher in Control beds compared to Late Water beds. This trend was consistent for both ‘Mullica Queen’ and ‘Stevens’, with Control treatments producing more berries per square foot than their paired Late Water beds. When averaged across sites, Control beds maintained a consistently higher berry count regardless of year. Variation in berry number was observed within individual bog pairs and cultivar, but the overall direction of the treatment effect remained the same for each cultivar.
3.5. Harvest and Fruit Quality Assessments
Inferential Analysis for the Year of Late Water
Fruit Rot: Percent rot was significantly higher in Late Water beds compared to Control beds during the year of treatment (Figure 10). The final linear mixed-effects model included Treatment, bog age, and total nitrogen as fixed effects, with location as a random effect. Based on the parameter estimates, Late Water increased percent rot by 13.48 percentage points, while bog age and total nitrogen were also significant, each increasing percent rot by 0.72% and 0.41% respectively. The model explained a substantial portion of the variability in percent rot, with a Conditional R² of 0.435 and a Marginal R² of 0.297. The model-estimated mean percent rot for Late Water beds was approximately 25.4%, compared to 12.0% in Control beds.
Yield: Yield was significantly reduced by the Late Water treatment. Yield was also significantly affected by cultivar and by the interaction between cultivar and Late Water. The final linear mixed-effects model included Treatment, Cultivar, and their interaction as fixed effects, with location as a random effect (Figure 11). Parameter estimates indicated that Late Water reduced yield by 13,157 kg/ha, and ‘Stevens’ yielded lower overall than ‘Mullica Queen’, with an estimated reduction of 23,392 kg/ha compared to ‘Mullica Queen’. The interaction between Late Water and ‘Stevens’ was positive, increasing the yield by 10,244 kg/ha. The model fit was strong, with a Conditional R² of 0.604 and a Marginal R² of 0.653, explaining approximately 60–65% of yield variance. Cultivar-specific means showed that for ‘Mullica Queen’, Late Water bogs produced an estimated yield of 31,417 kg/ha, compared to 44,574 kg/ha in Control bogs. For ‘Stevens’, yields under Late Water were 18,269 kg/ha compared to 21,182 kg/ha in Control bogs. The reduction in yield for ‘Mullica Queen’ under Late Water was significant, while the smaller decrease for ‘Stevens’ did not reach significance.
TAcy:
The final mixed-effects model for TAcy included Treatment, Cultivar, Total phosphorus, and Leaf Area Index (LAI) as fixed effects (Figure 12). The inclusion of Total phosphorus and LAI indicates that fruit color was associated with nutrient status and canopy density. Late Water had a positive effect on TAcy, increasing mean values by 3.34. Total phosphorus also had a significant positive effect, increasing TAcy by 1.50, while LAI had a negative effect, reducing TAcy by 1.62. Cultivar effects showed that ‘Stevens’ had significantly lower TAcy than ‘Mullica Queen’, with an estimated difference of –8.17. For ‘Mullica Queen’, Late Water berries had an average TAcy of approximately 33.0 compared to 29.7 in Control beds. For ‘Stevens’, Late Water berries averaged 24.8 compared to 21.5 in Control beds. The model explained a high proportion of variance, with a Conditional R² of 0.867 and a Marginal R² of 0.497.
Fruit Firmness: Late Water slightly reduced fruit firmness during the year of treatment (Figure 13). The mean firmness index for Control plots was approximately 673, while Late Water plots averaged about 653. The contrast between Control and Late Water was 19.6 (SE 9.81), and this difference was statistically significant (t = 1.99, p = 0.0484). Nitrogen also had a significant effect, reducing firmness by 1.8. The final mixed-effects model included Treatment and Total nitrogen as fixed effects with location as a random effect. The model had a Conditional R² of 0.744 and a Marginal R² of 0.133.
Inferential Analysis for the Year After Late Water (Lag Effect)
Lag effect on Fruit Rot: The final mixed-effects model for percent fruit rot in the year after Late Water included Treatment, Cultivar, and Fungicide applications as fixed effects, with location as a random effect. The model’s Conditional R² was 0.397 and Marginal R² was 0.334. Late Water did not differ from Control beds for percent rot in the year after treatment. Cultivar had a significant effect on percent rot, with ‘Stevens’ showing approximately 10% lower rot compared to ‘Mullica Queen’ (Figure 14)
Lag Effect on Yield: The final mixed-effects model for yield in the year after Late Water included Treatment, Cultivar, the Treatment × Cultivar interaction, and Total potassium as fixed effects, with location as a random effect. The model fit was strong, with a Conditional R² of 0.743 and a Marginal R² of 0.697. A significant positive lag effect of Late Water on yield was observed, with Late Water beds having higher yield than Control beds overall. For ‘Mullica Queen’, bogs with a Late Water history yielded an average of 53,125 compared to 43,648 in Control plots, an increase of 9,478 which was significant. For ‘Stevens’, yields were 27,478 with a Late Water history and 27,723 in Control plots, a difference of –245 which was not significant (Figure 15).
Lag Effect on TAcy: The final mixed-effects model for TAcy in the year after Late Water included Treatment, Bog age, and Total nitrogen as fixed effects. The Conditional R² was 0.849 and the Marginal R² was 0.281. There was no significant carryover effect of Late Water on berry anthocyanin content in the following year. Berries from plots that had Late Water the previous year had a mean TAcy of approximately 29.1 compared to 28.2 in plots without Late Water history. Bog age reduced TAcy by 0.64, while Total nitrogen increased TAcy by 0.16 (Figure 16).
Lag Effect on Fruit Firmness: The final mixed-effects model for fruit firmness in the year after Late Water included Treatment, Cultivar, Bog age, Total nitrogen, and Total potassium as fixed effects. The Conditional R² was 0.838 and the Marginal R² was 0.266. Fruit firmness was significantly lower in the beds that received Late Water the previous year compared to Control beds. Cultivar Stevens, nitrogen and potassium had positive effects in the year after late water while bog age had a negative effect. For ‘Mullica Queen’, mean firmness was approximately 503 in the late water beds and 518 in control beds plots without. For ‘Stevens’, mean firmness was 618 with a late beds and 633 with no history (Figure 17).
Bayesian Network Structure : Various factors considered as Nodes, edges in Bayesian network (Figure 18)
The conditional Gaussian Bayesian network identified three direct parent nodes for Percent Rot: Treatment, Total N, and Fungicide use. The edge from Treatment to Percent Rot had an estimated edge strength of approximately 1.0, indicating this connection was consistently present in the network across bootstrap samples. The edge from Fungicide to Percent Rot also had an estimated strength near 1.0, confirming its high stability in the learned structure. The edge from Total N to Percent Rot was weaker but still retained in the final structure, with an estimated edge strength of 0.59 (Figure 18, 19).
The structure for Yield included four direct parent nodes: Cultivar, Bog Age, Treatment, and Berry Number. The edges from Cultivar to Yield and from Treatment to Yield each had an estimated edge strength of approximately 1.0. The edge from Berry Number to Yield had an estimated edge strength of 0.99, indicating a strong and nearly deterministic relationship in the model. The edge from Bog Age to Yield had an estimated edge strength of 0.76. No direct edge was detected between Yield and Percent Rot in the final structure, indicating that the two outcomes were modeled independently with no direct causal pathway connecting them in the learned network (Figure 18, 19).
Percent Rot Parameter Estimates
The conditional Gaussian model for Percent Rot included an intercept estimate of 7.00. The effect of Treatment on Percent Rot was estimated at +7.93%. The effect of Total N was estimated at +0.13% for each one-unit increase in Total N. The effect of Fungicide was estimated at +0.41% per unit increase in the number of fungicide applications. The estimated parameters indicate that Percent Rot varied as a function of Treatment status, nitrogen level, and number of fungicide applications, with the baseline intercept providing the reference level for untreated fields with minimum nitrogen and fungicide input (Figure 18, 19, 20, 21).
Yield Parameter Estimates
The conditional Gaussian model for Yield produced an intercept estimate of 13,032 kg/ha. The coefficient for Cultivar was estimated at –14,909 kg/ha, indicating the effect of switching from Steven to Mullica Queen on expected yield. The coefficient for Treatment was estimated at –3,231 kg/ha. The effect of Bog Age was estimated at +380.8 kg/ha for each unit increase in bog age. The coefficient for Berry Number was estimated at +106.7 kg/ha for each unit increase in berry count. The network structure and associated parameter estimates indicate that Yield varied as a function of Cultivar selection, Treatment status, field age, and berry count, with the intercept representing the reference scenario (Figure 18, 19, 20, 21).
This multi-year, multi-site study provides one of the most comprehensive assessments to date of the immediate and residual (lag) effects of the Late Water (LW) spring flooding practice on cranberry yield, fruit quality, and plant physiological traits in southeastern Massachusetts. Our findings reveal that while LW can reduce early-season fungicide inputs and improve certain fruit quality traits, such as anthocyanin content (TAcy), it also presents trade-offs—particularly reduced yield and increased fruit rot during the year of application, especially for ‘Mullica Queen’. These outcomes are likely linked to lower carbohydrate reserves (TNSC), delayed bloom progression, and the application of nitrogen and fungicides in LW bogs at the same levels and timing as in control beds. Notably, fungicides were applied on nearly identical dates for both LW and control beds, despite the delayed bloom in LW sites, likely making fungicide applications prematurely timed for optimal fruit rot control in LW-treated bogs.
Importantly, our lag-year analysis showed that yield penalties associated with LW do not necessarily persist. In fact, ‘Mullica Queen’ demonstrated a significant positive yield response in the season following LW application, suggesting potential recovery or compensatory vine vigor.
Our integrated Bayesian network analysis clarified the complex relationships among management practices, physiological responses, and productivity outcomes. Cultivar, bog age, and LW treatment were identified as key drivers of yield, while berry number and reproductive investment (upright ratio) emerged as critical mediators. For fruit rot, nitrogen inputs and fungicide applications were strong predictors, with LW directly contributing to increased disease pressure during the year of application.
Overall, this research highlights the complex, context-dependent nature of LW effects. While LW holds potential to reduce chemical inputs and improve fruit color, its successful implementation requires cultivar-specific strategies and careful management. Our findings emphasize the need for a more nuanced, data-driven approach to LW decision-making—particularly in today’s regulatory and market environment that increasingly favors reduced pesticide use and higher fruit quality standards.
This study also surfaced several critical areas for future research. Although our observational design provided valuable insights into real-world grower practices, it limited our ability to control variables such as nitrogen input rates or fungicide scheduling. For example, applying standard nitrogen rates to LW beds—rather than reducing them as literature recommends—may have contributed to excessive vegetative growth, denser canopies, and increased humidity, all of which can promote fruit rot. Controlled trials using optimized management strategies (e.g., reduced nitrogen and adjusted fungicide inputs in LW beds) may yield more favorable LW outcomes, particularly with respect to balancing rot control and yield retention.
While these results provide growers with much-needed data and guidance to inform LW decisions, we emphasize that additional controlled studies are essential to further refine and improve the Late Water decision-making tool. Specifically, future research should incorporate a broader range of field sites, multiple years of data, and expanded treatment comparisons—including variable nitrogen application rates, differential fungicide frequencies, and timing adjustments aligned with bloom progression differences between LW and control beds. Such studies would enable a more comprehensive understanding of treatment interactions and their cumulative impacts on cranberry productivity and quality.
We strongly recommend continued development and implementation of a Late Water Decision Support Tool that integrates these multifaceted factors—such as cultivar, bog age, carbohydrate status, and seasonal weather conditions—to help growers make site-specific, cultivar-sensitive LW decisions that minimize risk while optimizing fruit quality and yield.
Education & Outreach Activities and Participation Summary
Educational activities:
Participation Summary:
Outreach Presentations:
- https://www.youtube.com/watch?v=_YvDCQn36i8 "200 Years of American Cranberry Domestication & Status of Fruit Rot Research". Presented for BioIngene Online Webinar.
- Leela Uppala_April 27, 2021 UMass Cranberry Station's Pesticide Safety Meeting "Cranberry Fungicide Options: A Review".
- Leela Uppala- November 19th, 2021 UMass Cranberry Station's Oversight Meeting.
- Leela Uppala- December 17th, 2021 Invited Seminar "Cranberry Disease Research" to University of Wisconsin.
- Leela Uppala, Salisu Sulley, Michael Nelson- April 8th, 2022. Late Water Advisory Board Meeting.
- Leela Uppala, Salisu Sulley, Michael Nelson- September 12th, 2022. Late Water Bogside Workshop.
-
Uppala, L. S., Nelson, M. F., and Salisu, S. 2022. Studies to identify critical criteria impacting late water outcome in Massachusetts cranberry production and develop a decision-making model. American Phytopathological Society Annual Meeting. August 6th-10th, Pittsburgh, PA.
- Leela Uppala, Salisu Sulley, Michael Nelson- September 7th, 2023. Late Water Bogside Workshop.
- Leela Uppala- November 17, 2023 UMass Cranberry Station's Oversight Meeting.
- Leela Uppala and Salisu Sulley, -January 30, 2024, UMass Cranberry Station's Annual Update meeting- Cranberry Fruit Rot: Research & Solutions
- Leela Uppala- November 22, 2024, UMass Cranberry Station's Oversight Meeting.
- Leela Uppala and Salisu Sulley, -January 28, 2025, UMass Cranberry Station's Annual Update meeting- Cranberry Fruit Rot Research Updates
Learning Outcomes
The current increased awareness of outcomes of late water flooding in new generation cranberry hybrids. Provided data driven guidance for late water use.