Water deficit can alter floral traits with cascading effects on flower-visitor interactions and plant fitness. Water stress induction can diminish productivity, directly resulting in lower flower production and consequently seed set. Changes in floral traits, such as floral scent or reward amount, may in turn alter pollinator visitations and behavior and consequently can reduce pollination services resulting in lower reproduction output. However, the relative contribution of this indirect in comparison to the direct effects of changes in seed set are not fully understood.

We manipulated water availability using rain-out shelters in a field experiment and measured effects on floral scent bouquet, morphology, phenology, flower-visitor interactions, pollination, and seed set. Plant individuals of Sinapis arvensis (Brassicaceae) were randomly assigned to one of three treatments: mean precipitation (= control), reduced mean precipitation, or drought period treatment.

Our results show that decreasing water availability lowers the number of flowers and seed set. This indicates a direct link between water stress and seed set, as seed mass increases with increasing flower number. The indirect link of water stress via floral traits, pollinator visits, and pollination has weaker effects on seed set. However, floral traits remain relatively stable under decreased water availability, whereas plant growth and flower abundance decrease, potentially in order to allow investment in more resources in fewer flowers to maintain pollination success. Thus, plants are able to compensate for water stress and can maintain floral trait expression, such as a stable scent emission and bouquet, to retain pollinator attraction.

These findings indicate that the direct link from water stress to seed set has a stronger impact on plants' reproductive success than the indirect link through altered floral trait expression and pollinator visits in a generalist plant species.

- manipulation of water availability on wild mustard (Sinapis arvensis) using rain-out shelters

- 3 treatments: well-watered, reduced watered, drought period; stem water potential was measured to conform plant's water stress and was used as the linear variable for further statistical analysis

- "Foral_traits"-file: following parameters were measured once in three freshly open flowers per plant from a low, middle, and high position to avoid position and age effects: stamen, style, and calyx length, petal length and width, flower display size, and nectar volume. On two flowers per plant, we collected one anther each in order to count pollen grains by means of a microscope (after 3 weeks under treatment). For pollen and nectar collection, nearly open flowers were covered with mesh bags the day before collection in order to prevent access by flower visitors. The inflorescence size (greatest expansion) was measured once on five inflorescences per plant. Floral height (height of the highest flower) was measured weekly with a folding yardstick; means values were included in the data file and were used for statistical analyses

- "Int_data"- file: plant-animal interactions were observed: each plant individual was observed daily; the number of visits by arthropods was recorded; the number of visits per day and per flower was calculated and used for further statistical analyses

- "Scent_data"-filechemical (= floral scent) data were analyzed using a thermal desorption system coupled with a GC-MS; compounds were analyzed and identified using GCMSsolution package; compounds in flowers were compared with those found on blank controls; the amount of compounds was estimated by comparing peak areas with area of a standard (Octadecane C18); absolute compound amounts included in the data file

- "Loggerdata"-file: raw data obtained from data loggers that were placed under each shelter; measured temperature and relative air humidity; one logger was placed next to the shelters and recorded temperature, relative air humidity, wind speed, photosynthetic active radiation PAR and air pressure

- "Soil_humi_temp"-file: raw data of weekly measured soil humidity and soil temperature next to each plant individual using avolumetrix water content sensor (Hydrosense II, Campbell Scientific) and a simple thermometer (DET3R, Voltcraft)

- "Plant_list"-file:  file with individual plant information

Data are saved in a common data exchange format ".csv." Statistical anlyses were made with open source software R studio.