Rewetting prolongs root growing season in minerotrophic peatlands and mitigates negative drought effects

We used minirhizotrons to monitor root phenology. Thereto, transparent tubes were installed at an angle of 45° in the soil to insert a root image scanner (CI-600 In-Situ Root Imager; CID Bio-science Inc., Camas, WA, USA), taking c. 350° scans (image size: 21.6 × 19.6 cm) of the tube-soil interface and thus roots at three depths (0–15, 15–30, and 30–45 cm). The aboveground part of the tubes was wrapped with mirror foil to reduce thermal differences, taped and covered with a cap to exclude light from the tubes. Installation of the tubes took place in mid-August 2017 and measurements began in April 2018. As our study sites are very dynamic, fast-growing systems, we consider this time sufficient for recovery from the disturbance caused by the installation. Three minirhizotrons were installed within a distance of 1 m within each of the five plots at each of the respective sites, but one tube was damaged during the study, resulting in a total of 89 minirhizotrons (six sites × five plots × three tubes). In the rewetted coastal fen site (CW), images could only be taken down to a depth of 30 cm for 10 tubes, due to hard mineral soil limiting the depth of minirhizotrons. We measured biweekly until 15th October 2018 and then monthly until 5th December 2018, resulting in 16 image sampling events. We processed the sample images with our newly developed automated RootDetector, which outputs binary images in which white pixels represent detected root objects and black pixels represent background or non-root objects. We applied a topology-preserving thinning algorithm as implemented by the skeletonize function in scikit-image v. 0.17.1 to reduce the detected root objects to one pixel wide lines. The output was converted into root length in mm cm-2, given that root length is stronger related to root functionality (e.g., nutrient uptake) than mass. Soil temperature data was collected at 15-minute intervals at 5 cm and 15 cm depth by six loggers per site (HOBO, Onset Computer Corporation, Bourne, MA, USA). Groundwater table relative to soil surface was recorded at 15-minute intervals in a slotted PVC pipe using a CS456 pressure transducer connected via an SDI-12 sensor to a CR1000 data logger (Campbell Scientific Ltd., Bremen, Germany) at the alder forest sites, by Dipper-PT loggers for the two percolation fen sites and the drained coastal fen, and by a Baro-Dipper and a Dipper-APT logger (SEBA Hydrometrie GmbH & Co. KG, Kaufbeuren, Germany) for the rewetted coastal fen site. Gaps in water table data recording for the rewetted alder forest site between 23rd August and 27th October 2018 resulted from water tables below - 70 cm, which exceeded the reach of the groundwater pipe at this site. We used R version 4.0.2 (R Core Team, 2020) for all statistical analyses and visualizations (R package ggplot2, version 3.2.1).