Water depths, as measured in the middle of the fields, were maintained between 0.3 m to 0.5 m for all years. Inlet structures were fitted with 3-mm mesh screens to permit water inflow and prevent egress of stocked salmon. Outlet structures were fitted with 3-mm mesh screens in the 2013 and 2016 experiments. However, in 2014 and 2015, outlet structures were left open with a 5-cm diameter hole drilled in the middle of a 3.8-cm × 14-cm board and placed near the top of the water level in the rice box to investigate volitional out migration patterns of the stocked salmon. Each outlet structure was fitted with a live car trap placed in the outlet canal, which allowed for collection of all exiting fish. In 2014 and 2015, live cars were checked daily for the duration of the experiments to enumerate the number of emigrating salmon. In past experiments we observed a tendency for a portion of hatchery fish to “scatter” upon initial release into floodplain fields. This behavior reliably abated after several days as fish acclimated to new conditions. For this reason, downstream exiting fish were restocked back to the inlet side of the fields for the first week of 2014. In 2015, fish were similarly restocked for two weeks.Substrate type– 2013. After harvest, vertical grow system rice farmers typically treat the residual rice straw remaining in the fields using one of several methods; thus an important question was whether differences in treatment of rice straw created different outcomes for rearing fish.
Nine fields were randomly assigned to one of three post-harvest substrate treatments: rice stubble, disced, or fallow. The rice stubble substrate treatment consisted of standing stalks that remained after rice plants were cut for harvest using a rice harvesting combine tractor. The disced treatment consisted of plowing rice straw into the soil, a practice farmers use to promote stubble decomposition. The fallow habitat had not been planted with rice during the previous growing season but instead consisted of weedy herbaceous vegetation that voluntarily colonized the fields during the growing season and was left standing during the experiment. More details on the 2013 experimental design can be found in publications by our colleagues. Depth refugia– 2014. Avian predation on fish in aquaculture fields is a well-known problem. Avian predation has the potential to be a significant source of fish mortality in winter-flooded rice fields as California’s Central Valley is positioned directly within the winter habitat of diverse bird populations in the Pacific flyway. We evaluated whether trenching could provide depth refuge as a potential method for reducing avian predation on fish in winter-flooded rice fields. In 2014, nine fields all with a disced substrate, were randomly assigned to one of three treatments: three fields were assigned no perimeter trench, three were assigned a 0.5 m deep perimeter trench, and three were assigned a 1.0 m deep perimeter trench. All trenches were constructed on the north and east sides of the fields running continuously from the inlet structure in the northwest corner to the drain structure in the southeast corner. All trenches were approximately 1.0 m wide with the outermost edges of the trench spaced approximately 1.0m from the exterior levee surrounding the field.
We created this spacing specifically so depth refuges were outside the striking distance of wading birds such as herons and egrets which frequent the shallow water of the perimeter levees. Survival data for three fields was excluded from the analysis due to loss of containment on the inlet side of three fields during the last week of the experiment allowing fish to escape upstream into the inlet canal. Ancillary effects of the trench treatments on field drainage efficiency and volitional migration of fish were also investigated. Drainage practices– 2015. We wanted to know if we could create artificial hydrologic cues to trigger fish out-migration from fields. To investigate drainage practice effects on fish survival, the nine fields were randomly assigned one of three draining treatments: 1) fast drain, where inlet water was cut off and outlet boards were removed rapidly, resulting in the water draining off the fields in a single day; 2) slow drain with inflow, where water levels were lowered by 5 cm per day at the outlet while inflow was maintained through a mesh screen; and 3) slow drain without inflow, where water levels were lowered 5 cm per day at the outlet and inflow was cut off by boarding up the inlet structure. The drainage duration for both slow drain procedures lasted for 10 days with daily outmigration of salmon measured in the outlet traps. All nine experimental fields had a rice stubble substrate following the rice harvest in fall 2014, and a 0.5 m deep perimeter trench was constructed in all fields connecting the inlet and outlet structures running along the north and east sides of the fields. The trenches were approximately 1.0 m wide and spaced 1.0 m infield from perimeter levees. Survival through time– 2016.
To examine in-field survivorship of juvenile salmon through time, fish were stocked in six of the nine flooded experimental fields. During each of following six weeks, one randomly selected field was drained using the fast drain procedure, detailed in the 2015 experiment. All fields had fallow substrate as described in the 2013 experiment and 0.5 m deep trenches as described in the 2015 experiment. An impending bypass flood event near the end of the study forced the drainage of the last field 4 days earlier than scheduled. In-field water quality. Across all years and fields, we recorded continuous water temperatures in 15-min intervals using HOBO U22 temperature loggers anchored in a fixed vertical position on a metal t-post approximately 10 cm above the substrate in the middle of each field as well as trench substrate for a representative set of treatments when applicable. Localized temperature refugia in the trenches was evaluated in its capacity to create thermal buffering by comparing daily maximum water temperatures in the bottom of a trench to those in the middle of the field. Analysis of other physical water quality parameters, nutrient loading, and primary productivity in these experimental rice fields can be found in publications by our colleagues. Zooplankton abundance. Throughout all years, a randomly stratified subset of three fields was sampled for zooplankton weekly except in 2013 where all nine fields were sampled weekly. A 30-cm diameter 150-μm mesh zooplankton net was thrown 5 m and retrieved through the water column four times, once in each cardinal direction. In 2013, benthic macroinvertebrates were sampled separately using benthic sweeps, but due to high sedimentation, high spatial and temporal sample replication, and low overall contribution to the invertebrate community, the additional processing was deemed unnecessary in subsequent years. Furthermore, the zooplankton tow method is effective for assessing pelagic zooplankton and macroinvertebrate community assemblages while improving sample processing efficiency since it avoids the heavy sedimentation associated with benthic sweeps on wetland substrates. Additionally, we also relied on the stomach contents of in-field salmon to better inform the assemblage of macroinvertebrates present in the floodplain food web and their contribution to the diet of in-field salmon . Sampling location in each test plot was determined randomly via a selection of random x and y distances from a random number table. All samples were preserved in a solution of 95% ethanol. Organisms were identified with the aid of a dissecting microscope at four times magnification to the lowest taxonomic level possible using several widely recognized keys. Abundance estimates were calculated from homogenized subsamples of known volume and extrapolated to the volume sampled during the initial net throws. Salmon stomach contents. A random sub-sample of in-field salmon captured during weekly sampling with 4.8-mm mesh seine and sequential field draining were sacrificed, pipp racking transported on ice, and stored in a freezer at -22˚C. A total of 532 salmon stomachs were dissected using a dissecting microscope at four times magnification. Prey items were enumerated, but due to variable decomposition, prey item identification in the stomachs was limited to taxonomic order. Overall salmon survival and growth. Estimates of initially stocked salmon in each field were calculated by establishing a fish per kilogram ratio and multiplying by the total weight applied to each field, except in 2016 where the overall number of stocked fish was sufficiently low to count individually.
Stocking density was calculated by dividing the estimate of initially stocked salmon by the field area . Fish lethally sampled for stomach content analysis during weekly sampling were subtracted from the initial stocking estimate. Total salmon survival in each field was cumulatively enumerated in the outlet live car traps except during the restocking phase of 2014 and 2015 when volitionally emigrating fish were restocked to the inlet side of the fields. During field drainage, seines were used to collect stranded fish out of standing water and these fish were added to the cumulative survival count from the outlet live cars with the recovery method recorded. Survival in 2015 was calculated from only the fast drain treatment fields since the drawn out drainage methods were not comparable to drainage methods in other years. Prior to stocking in each year, mean initial fork length and wet weight were calculated from a random sample of 30 live fish measured to the nearest millimeter and weighed to the nearest hundredth of a gram with an Ohaus Scout Pro SP202 scale . For 2013–2015, we conducted weekly in-field fish sampling with a 4.8 mm mesh seine to capture a target of 30 fish per treatment, with the fork length and wet weight measured. In 2016, fish size data were collected from a random sample of 30 fish in out-migrant traps as individual fields were drained weekly.Percent survival for each field was calculated as the total number of recovered fish divided by number of initially stocked fish, times 100. Analysis of covariance was used to test for interaction effects between field substrate treatment and time which would indicate treatment effects on salmon growth rates. In this model, fork length was the dependent variable with field substrate, day of the experiment and an interaction term as the independent variables. When the assumptions of normality and homogeneity of variance were satisfied, as tested by the Shapiro-Wilk and Levene tests respectively, a one-way analysis of variance was used to test for significant differences in survival due to field drainage treatments. A post hoc Tukey honestly significant differences test was used to test all pairwise comparisons of field drainage practices. When the assumptions of normality and/or homogeneity of variance were not satisfied, non-parametric Kruskal-Wallis analysis was used to test for significant differences in survival and daily volitional outmigration due to field trench depth treatments and to test for differences in overall mean total zooplankton densities between years and substrate types. A post hoc Dunn’s test was used to test all pairwise comparisons of daily volitional outmigration due to field trench depth treatments. Linear regression was used to estimate apparent growth rates and to examine the relationship between salmon survival and day of the experiment . Linear regression was also used to evaluate the degree of within-field thermal refugia via the relationship between daily maximum water temperature differences in the trenches and daily maximum water temperature in the middle of the field . Statistical significance was declared at an α = 0.05 level. All analyses were conducted in R v3.6.1.Apparent fork length growth rate for juvenile salmon did not differ significantly between treatments . The slopes from individual linear regressions of fork length predicted by day for each treatment resulted in estimated apparent growth rates of 1.01 mm d-1 for the stubble treatment, 0.99 mm d-1 for the disced treatment, and 0.95 mm d-1 for the fallow treatment. As previously published, found no statistical difference between total abundance of zooplankton between treatments, but did find high overall abundance and a trend of increasing zooplankton over experiment duration. Across all samples, cladocera were the most abundant group of zooplankton, making up over 50% of the total zooplankton assemblage. Cladoceran zooplankton was the most common prey item found in juvenile salmon stomach contents as this taxon comprised on average 94.0% ± 1.0% SE of the diet composition across all treatments. Chironomid midges were the second most common prey item and comprised an average of 4.8% ± 1.0% SE of the diets.