Best management practices that promote sustainable crop pollination: the role of crop rotations and tillage depth

Study system:

The study was carried out in Yolo County, California. There are no wild species of Cucurbita found in this county. Squash bees, therefore, depend on Cucurbita species grown as crops. Squash bees are univoltine, solitary bees that nest preferentially underneath or near Cucurbita vines (Julier and Roulston 2009). Squash bee offspring produced in one summer overwinter underground and emerge the following summer (Hurd et al. 1974). Individuals that survive tilling must locate squash, pumpkin, or gourd fields within their dispersal range from which to collect pollen to provision nests.

Tilling practices (Objective 1):

In summer 2012 we set up twenty 3m by 3m by 1.8m cages over three varieties of squash planted earlier in the season (Fig 1). We then released two female and two male squash bees into each sealed cage to establish artificial nesting aggregations. We monitored cages to count the number of nests observed and to ensure that bees were provisioning nests with pollen. In late fall 2012, after all adult bees had died, we removed cages. We then randomly selected ten cage locations to apply a tillage treatment. The tillage treatment included disking to 15.2 cm, ripping to 40.6 cm, and then disking again at 15.2 cm. In spring 2013 we placed the cages back over their previous locations. We collected all emerging offspring in blue vane traps (Springstar Inc, Woodinville, Washington, USA) filled with soapy water. In collaboration with Mathew Meisner (University of California, Davis), we used Bayesian statistics to quantify the relationship between tilling and the number of emerging offspring. We considered tillage to have a strong effect on the offspring emergence if the 95% highest posterior density (HPD) interval, a Bayesian analogue of a confidence interval, for the tillage parameter did not overlap zero (Severini 1991). Alternatively, if the HPD interval slightly overlapped zero, we used information about the posterior distribution to interpret the results. We fitted an overdispersed Poisson mixed model to quantify the effect of tillage on the number of emerging squash bee offspring trapped per cage. Treatment, number of nests per cage, and Cucurbita variety were included as fixed effects. Cage ID was included as a subject level random effect to account for overdispersion.

Crop rotations (Objective 2):

We sampled 65 squash fields in California that differed in their connectivity to surrounding squash fields over time. At each field we estimated the density of squash bees along two to four 50 square meter transects, for a total of 100-200 square meters per field. For each transect we observed and recorded all squash bees visiting squash flowers during a 10 minute interval. Prior to analysis we subsampled the data such that each site only had data from two transects (e.g. 100 square meters). Fields were sampled under similar weather conditions. Voucher specimens were collected at the end of each sample round to verify observational identifications. Sampling occurred between July 3, 2012 and September 4, 2012. Field size, in square meters, and distance between squash fields within and across years was calculated in ArcGIS using GIS data provided by the California Department of Pesticide Regulation and verified through ground surveys. Within-year connectivity was measured as the Euclidian distance between the focal field and all surrounding fields. We assumed that squash fields surrounding the focal field that fell within the foraging range of squash bees (e.g. 1600m based on intertegular distance) contributed more bees and were weighted accordingly (Greenleaf et al. 2007, Winfree et al. 2006). To measure between year connectivity we calculated the same measure but looked at distance between the focal field and the location of all squash fields grown in the previous year (e.g. 2011). We fit a generalized linear model with a negative binomial distribution to account for overdispersion in the data. Our response variable was squash bee density per field in 2012. Covariates included Julian date the field was sampled and either within-year or between-year connectivity. We used Akaike’s Information Criteria (AIC) to determine which model best explained the data ( AIC <2, Burnham & Anderson 2002).

Modelling P. pruinosa populations (Objectives 3-4):

In collaboration with Matt Loiacono and Eric Lonsdorf (both at Franklin and Marshall College) we developed a spatially explicit population model, modelling female squash bees only. The model assumed that squash bee population growth is a function of overwintering and tilling survival, dispersal, and number of eggs laid. We assumed that overwintering survival and number of eggs laid was constant across squash fields, but that tilling survival and dispersal survival varied across squash fields. We assumed that if the squash bee survived tilling and emerges, she will need to find a new squash field to nest in. Squash bee survival is therefore dependent on a field being within 1,600 meters of where the squash bee emerged. This distance estimate is based on squash bee intertegular distance (Greenleaf et al. 2007). We applied the model to a landscape that consisted of 540 squash fields using locations of squash fields over a five year period. Squash field locations were identified using GIS data provided by the California Department of Pesticide Regulation. The flowering pattern was repeated for any number of years the model ran for. We ran an initial 60 year simulation to evaluate population growth. Parameter values were chosen to ensure that the model ran stably. We then ran the model exploring the impact of tilling on the total population after a 30 year simulation. We ran 50 iterations of this simulation under different tilling conditions across the landscape, e.g. 0, 20%, 40%, 60%, 80%, or 100% of fields experienced tilling. In order to determine the impact of squash field connectivity over time, we ran a 30 year simulation without tilling. We quantified the number of squash bees per field at three different time intervals (after 5 years, 15 years, and 30 years) to determine if areas of high connectivity over time also had more squash bees over time.  

• Fig 1 – Tilling