Precipitation seasonal distribution experiment

Earth system models have projected an intensification of global hydrological cycles in the future. The changes in global water circulation are predicted to enhance the inter- and intra-annual variability of precipitation, especially the seasonal distribution of precipitation. Given that precipitation/water availability is one of the dominant limiting resources for plant photosynthesis, growth and productivity and that terrestrial plants have differential water requirement during different growth stages, seasonal redistribution of precipitation, such as advanced or delayed precipitation peaks in the growing season, may play a more important role than precipitation amount in regulating ecosystem structure and function. In order to investigate effect of precipitation seasonal redistribution on ecosystem structure and function, a field experiment simulating advanced and/or delayed precipitation peaks without altering total precipitation amount in the growing season was conducted.

The semiarid grasslands in this area represent the typical vegetation across the Eurasian content and have profound influences on biodiversity and ecosystem service at regional scale. Over the past half century (1954-2017), mean monthly precipitation over the growing season (May to October) in the local area showed a unimodal pattern with the precipitation peak occurring in the middle growing season (July and August; Fig 1), but the monthly precipitation showed declining trends in both July and August and increasing trends in both June and September (Fig. 1). Thus, we tentatively advanced and/or delayed precipitation peaks in the growing season by switching monthly precipitation between June and July, between August and September, and both (Fig. 2).

This study used a randomized complete block design with four treatments and six replicates for each treatment. Six permanent 14 × 14 m² rainout shelters (3.5 m high at the roof and 2.0 m at the edge) with steel frames and covered with clear polyethylene roofs were built in late spring 2013 to control precipitation inputs from June to September each year. Each rainout shelter covered one block. Four 5 × 5 m² plots were arranged in each of the six blocks. The distance between any two adjacent blocks was 4 m and the distance between any two adjacent plots was 2 m in the same block. The 4 plots in each block were randomly assigned to four treatments: control (C, natural precipitation, Fig. 1a), advanced growing-season precipitation peak (AP, shifting monthly precipitation from July to June and from June to July, Fig. 1b), delayed growing-season precipitation peak (DP, shifting monthly precipitation from August to September and from September to August, Fig. 1c), and the combination of AP and DP (ADP, Fig. 1d). The total precipitation amount was kept constant for all the four treatments. In addition, there was a 4 × 4 m² core area with 0.5-m buffer zone to the edge of each plot and all the measurements were conducted in the core area to reduce the edge effect.

The polyethylene roofs were added by the end of May and removed by the end of September to allow all the plots to receive natural precipitation from October to May in the next year. During the time periods from June to September, natural precipitation was intercepted by the rainout shelters, collected, stored in 8 black plastic buckets, and applied manually back to the plots assigned to the 4 treatments. The years with both rainfall size and frequency from June to September close to the long-term means were selected to simulate natural precipitation in the local area.

Figure 1. Temporal trends of precipitation in June, July, August, and September from 1954 to 2017 and from 1960 to 2017.

Figure 2. Precipitation distribution under different treatments. (a) Control; (b) Advanced growing-season precipitation peak; (c) Delayed growing-season precipitation peak; (d) Both advanced and Delayed growing-season precipitation peaks.

Figure 3. Plot layout of the experiment.