Fifty years of recorded hillslope runoff on seasonally-frozen ground: The Swift Current, Saskatchewan, Canada dataset

. Long records of hillslope runoff and nutrient concentrations are rare—on seasonally-frozen ground they are almost non-existent. The 10 Swift Current hillslopes at the Swift Current Research and Development Centre on the Canadian Prairies provide such a long-term hydrological record. Runoff, runoff nutrient concentration, snowpack depth, density and water equivalent, soil moisture, and soil nutrient concentration were monitored on the three 5 ha hillslopes over a 50-year period (1962 – 2011). Digital elevation data are available for the three hillslopes at a 2 m horizontal resolution, and, for one of the hillslopes (Hillslope 2), at a 0.25 m horizontal resolution. Runoff from the hillslopes was generated episodically during snowmelt and occasional rainfall events. Hillslope runoff was measured with a 0.61 m H-flume. Daily runoff nutrient 15 concentration data are available for nitrate-N (March 1971 – April 2011), ammoniacal-N (February 1996 – April 2011), and phosphate-P (March-April 1971; June 1991 – April 2011). Snowpack data (snowpack depth, density and water equivalent) were determined via manual snow surveys carried out several times each winter, between January and March, between 1965 and 2011. Gravimetric soil moisture content was measured in October and April each year between 1971 and 2011 at five depth intervals (0-15, 15-30, 30-60, 60-90, and 90-120 cm) at nine points on each hillslope. We provide these hillslope data in two publically-available repositories: 1) 1962-2011 data on runoff, runoff nutrients, snowpack, soil 20 moisture, soil nutrients, and crop and tillage practices at https://doi.org/10.23684/hhn5-rz52; and 2) digital elevation data at https://doi.org/10.20383/101.0117 ( Coles et al., 2018b). Complete climate data recorded at a Environment and Climate Change Canada meteorological station located 390 m from the three hillslopes are publically-available at http://climate.weather.gc.ca/. resolution. Data from the three hillslopes and the meteorological station are now available online. This rich dataset is a valuable source for hydrological response analysis and model formulation and calibration, within the context of climate and land management change.

Long-term datasets that chronicle hydrological processes within the context of climate and land management change are rare. This is especially true for cold regions with seasonally-frozen ground and where remoteness and inclement weather limit continuous measurements. Long-term experimental datasets that do exist in such regions are diminishing quickly (Laudon et al., 2017). For hillslope-scale long-term experiments over seasonally-frozen ground, such records are particularly uncommon. Here we present an unusually-long 50-year record for a set of three monitored agricultural hillslopes in southwest Saskatchewan, Canada that includes runoff, runoff nutrients, snowpack, soil moisture, and soil nutrients for a 5 50-year period, during which the hillslopes underwent controlled, comparative changes in agricultural practices.
The hillslope-scale dataset is at the spatial scale intermediate between more traditional long-term catchment-scale datasets (e.g. Water Survey of Canada (WSC) Historical Data, the United States Geological Survey (USGS) National Streamflow Information Program, or the UK National River Flow Archive) and point-scale datasets (e.g. TERENO-SOILCan network (Pütz et al., 2016) or the Reynold's Creek soil lysimeter network 10 (Seyfried et al., 2001)). As such, hillslope-scale measurements of precipitation and runoff, together with internal measurements of soil moisture, snowpack, and soil nutrients offer an opportunity to understand mechanistically the links between climate trends in rainfall and snowmelt inputs and runoff outputs. Further, in terms of land use change in dry agroecosystems, they offer an opportunity to understand how tillage, seeding, and crop rotational practices affect the hydrology of agricultural hillslopes, as well as nutrient delivery to downstream water bodies.

15
The collation of our long-term hillslope database has been motivated largely by the importance of water management and sustainability for dryland agriculture on seasonally-frozen, snowmelt-dominated terrain. This paper reports on an archive of downloadable data from an experimental, agriculturally-managed hillslope site at Swift Current, Saskatchewan. This dataset documents the characteristics more broadly of the semiarid portions of the northern Great Plains of North America. It serves as a resource for long-term analyses and model formulation, calibration, and testing of land management and climate change effects on agricultural hillslopes.

2 The Swift Current hillslopes
The three hillslopes (also referred to as 'watersheds' in the dataset) are located at South Farm (50°15'53" N 107°43'53" W), an agricultural research site of Swift Current Research and Development Centre of Agriculture and Agri-food Canada, approximately 5 km southeast of Swift Current, Saskatchewan (Figure 1). The hillslopes adjoin each other and are rectangular in shape, with areas of 4.25 ha (Hillslope 1), 4.66 ha (Hillslope 2), and 4.86 ha (Hillslope 3). The hillslopes are approximately 150 m wide in the east-west direction and 300-320 m long in the north-Hydrological observations at the hillslopes began in 1962 to address the effects of agricultural land management practices on runoff water supply and quality, chemical transport, and soil erodibility.
The hillslopes have undulating topography and slope towards the north-northwest with gradients of 1-4%. Digital elevation models (DEMs), obtained using a Leica Viva GS15 from 17-18 April 2012, are available for all three hillslopes at a 2 m horizontal resolution (Coles et al., 2018b).

5
These DEMs were obtained by mounting the Leica Viva GS15 rover to a side-by-side vehicle at a height of 2 m above ground surface. The vehicle was driven methodically up and down each hillslope, achieving full coverage of each hillslope (Coles et al., 2018b). The GPS was set to automatically collect points every two meters and at instances when the elevation change exceeded 0.15 meters (Coles et al., 2018b). The survey was referenced to a permanent Agriculture and Agri-food Canada benchmark located near to the hillslopes, to give elevations in meters above sea level. Location data for these DEMs are in the projected coordinate system UTM Zone 13N with GRS 1980 datum. Two additional DEMs, 10 obtained using an Optech ILRIS-LR Terrestrial Laser Scanner in summer 2014, are available for Hillslope 2 at a 0.25 m horizontal resolution (Coles et al., 2018b). These two finer DEMs of Hillslope 2 were obtained from surveys on 7 July 2014 and 24 September 2014, before and after a simulated seeding of the hillslope. They therefore capture the surface micro-topography of a tilled hillslope with random soil clod undulations, as well as a seeding-induced ridge-and-furrow topography. These DEMs were obtained by mounting the laser scanner at the top of a 15 ft scaffolding tower located on the east margin of Hillslope 2, and also from the top of a vehicle parked in the northeast corner of the hillslope. The

25
The hillslopes have been predominantly under an annual rotation of wheat (Triticum aestivum L.) and fallow (Table 1). This two-crop rotation is with the exception of: a period (1977)(1978)(1979)(1980) of grass (Psathyrostachys juncea (Fisch.) Nevski) and a period (1982)(1983)(1984)(1985) of annual wheat on Hillslopes 1 and 2; an annual rotation (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) of wheat and legume green manure (Lathyrus sativus L.) without use of herbicides or mineral fertilizers on Hillslope 1; and an annual rotation (2004-2011) of wheat and pulses (lentils and peas; Lens culinaris L. and Pisum sativum L., respectively) on Hillslope 2. The hillslopes have largely been under conventional tillage practice with a heavy-duty cultivator, with the exception of the period 1993-2011 when Hillslope 2 was switched to no tillage practice, with weed control entirely with herbicides. Unlike the other two hillslopes, Hillslope 3 has had a consistent two-crop rotation and consistent tillage management since 1962.   (2018) used single melt season intensive data collection to quantify the spatial patterns of hydrological variables and the generation of snowmelt-runoff 20 connectivity over low-angled terrain. Finally, climate change studies at this site have shown warming winter and spring temperatures, decreasing snowfall amounts and a resultant decrease in snowmelt-runoff amounts, earlier spring runoff timing, and an increase in summer rainfall amounts but no change in rainfall-driven runoff (Cutforth et al., 1999;Coles et al., 2017).

Available data series
The available data are summarized in Table 2 and detailed below.

Runoff
Runoff from the hillslopes is non-continuous, and generated during snowmelt and occasionally during heavy rainfall events. Figure 2 shows the 50-year runoff record for one hillslope (Hillslope 2). Of the 50 years on record, 46 of those years saw measurable flow during spring snowmelt (on at least one of the hillslopes) and 28 years saw measureable hillslope-scale runoff during non-melt rainfall-runoff events (also on at least one of the hillslopes). Because the flumes are so small, the minimum measureable instantaneous flow through the flume is 0.07 L. Assuming a reliable 5 reading would need 30 seconds at that flow rate, this translates to a minimum measureable daily flow of 0.000049 mm/day (Hillslope 1), 0.000045 (Hillslope 2) and 0.000043 mm/day (Hillslope 3). The dataset includes data on daily runoff (mm) and daily peak flow (L s -1 ). Any missing data are shown by an 'NA' and also flagged with an 'm'.
The raised grassed berms prevent runoff from being transferred between the hillslopes. Runoff from the hillslopes is routed through a 0.61 m H-10 flume (Bos, 1989) at the downstream outlet (north-northeast corner) of each hillslope. The flume state was measured between January 1962 and December 2011 using a Stevens (Portland, Oregon, USA) water level chart recorder in the stilling well of each flume. Since 1994, water level recorder shaft position encoder (Belfort Instrument Co., Baltimore, Maryland, USA) was the primary method to record water level. Daily cumulative runoff amounts (mm) were calculated from these water level measurements using a standard H-flume rating curve (Bos, 1989). The H-flumes were heated in cold weather to prevent icing. Prior to 1993, the flumes were in the open and heated with propane-fueled heaters under 15 the flumes. After 1993, the flume sides and bottoms were electrically heated with resistance heaters and the flumes were in a small building that was also electrically heated ( Figure 3). There is no runoff data from March 1969 to November 1970 as the flumes were not monitored. A heavy rainfall event on 14 June 1964 caused flow rates to exceed the flume capacity. Runoff during this event is reported as 'NA' in the dataset, but total daily runoff was estimated to be 72 mm with a peak flow of 60 mm h -1 (McConkey et al., 1997). Table 3

Runoff nutrient concentrations
Nutrient concentrations in the runoff were measured on a daily basis during runoff events, from runoff samples taken from the flume at the downstream outlet of each hillslope. Concentration data (in mg L -1 ) are available for nitrate- N (1971-2011), ammoniacal-N (1996-2011), and orthophosphate-P (1971; 1991-2011).

25
Prior to 1993, on days with appreciable runoff, water samples were collected manually in 0.5-L glass containers at mid-morning (10:00 h ± 30 min) and mid-afternoon (15:00 h ± 30 min). From 1993 onward, 0.5 L samples were collected using an automated water sampler (ISCO 3700 Portable Sampler, Isco, Inc.), previously described by Cessna et al. (2013). The collected water samples were filtered (No. 42,Whatman International filter papers, Maidstone, England) and then analyzed for dissolved NO 3 -N and orthophosphate-P according to the analytical procedure of Hamm et al. (1970) and for ammonium-nitrogen (NH 4 -N) according the analytical procedure of Gentry and Willis (1988). The runoff nutrient concentration data are summarised in Table 4.

5
Snowpack data were determined for each hillslope by manual snow surveys each year between 1965 and 2011. At nine points on each hillslope (as shown in Figure 1), an average snow depth (cm) was determined from multiple locations in the close vicinity of the point with a graduated rod and density (g cm -3 ) was measured once per point as the mass of snow of measured depth taken in in a 7.5-cm diameter core; the mass was determined in the field with a calibrated spring scale. The snow water equivalent (SWE; cm) was calculated from average snow depth multiplied by density. Snow surveys were carried out several times each winter, between January and March. Hillslope-averaged snow depth, density, and 10 SWE were calculated from the nine points for each survey. These data are summarised in Table 5.

Soil moisture
Gravimetric soil moisture content (water fraction by volume of soil) was measured twice each year between 1971 and 2011: in fall prior to freezeup (sometime in September-November), and in spring following snowmelt (sometime in April-May). The measurements were taken at the same nine points on each hillslope at which snow characteristics were measured (Figure 1), at five depth intervals per point: 0-15 cm, 15-30 cm, 30-60 15 cm, 60-90 cm, and 90-120 cm. The soil moisture was measured from a subsample of the entire mixed interval and reported for the mid-point of the interval. These gravimetric soil moisture contents were converted to volumetric soil moisture using average bulk density across the hillslopes derived from occasional measurements from mass of soil taken in cores of a known volume and depth. These bulk densities were: 1.26 g cm -3 for 0-15 cm, 1.29 g cm -3 for 15-30 cm, 1.39 g cm -3 for 30-60 cm, 1.54 g cm -3 for 60-90 cm, and 1.63 g cm -3 for 90-120 cm. Hillslope-averaged soil moisture at each depth was calculated from the nine points. Soil moisture data in the dataset are reported in cm of water for each soil profile depth 20 interval. A summary of the soil moisture aggregated over the entire 0-120 cm soil profile is presented in Table 6.

Soil nutrient concentrations
A subsample from the same samples collected for soil moisture in fall and spring was air-dried and analyzed for NO 3 -N (1970-1992; 1994-2010), bicarbonate-extractable (Olsen) P (1970-1992; 1994-2010), and ammoniacal-N (1970-1992) (Hamm et al., 1970). Data are reported in the dataset in units of concentration (mg L -1 ) and mass (kg ha -1 ). A summary of the fall soil nutrient concentration data is presented in Table 7.

Meteorology
To complement the hillslope data, meteorological data are available from the Environment and Climate Change Canada meteorological station (station name: Swift Current CDA; Climate ID: 4028060; WMO ID: 71446 and downloadable at http://climate.weather.gc.ca/), 390 m south of the southwest corner of Hillslope 1 and within 1 km of all three hillslopes. These data include daily (1962-present) precipitation (snowfall and rainfall), temperature, wind speed and direction, and snow depth, and hourly (1995-present) temperature, wind speed and direction, and relative 5 humidity.

Data availability
Data from the three Swift Current hillslopes are available at: https://doi.org/10.23684/hhn5-rz52. This depository includes the runoff, runoff nutrient concentrations, snowpack, soil moisture, soil nutrient concentrations, and crop and tillage data. Herbicide and sediment concentration data are not included in this dataset. We have also made available the digital elevation data for the site, which can be accessed at: