High-latitude conditions in northern Europe are characterised by short growing seasons (May–August) and long dormant seasons. Alternating mild and freezing conditions lead to variable snow accumulation–melt cycles affecting runoff generation, and consequently the loss of nutrients and sediments from agricultural fields. We assessed water balance in two subsurface drained clayey agricultural fields of different slopes (1% and 5%) in southern Finland to discern changes between mild and cold winters. The water balances of the two field sections were produced with a spatially distributed 3D hydrological model. Simulated snow water equivalent (SWE), drain discharge, tillage layer runoff and groundwater outflow from a 7-year period were examined during the dormant seasons (September–April) in relation to the North Atlantic Oscillation (NAO) index, which characterises phases related to mild and cold winters in northern Europe. Mild periods (positive NAO) were associated with more frequent runoff events, which were sustained throughout mild winters with lower SWE and shorter time of snow cover. Understanding and quantifying the water balance through periods of different weather patterns is essential as climate change is projected to increase the occurrence of positive NAO phases challenging the control of nutrient and sediment losses from agricultural fields.
INTRODUCTION
Northern agricultural areas are characterised by short growing seasons (May–August) with relatively high levels of summer evapotranspiration and long dormant seasons with low evapotranspiration due to decreased net radiation and cold temperatures. Precipitation and snowmelt drive runoff generation, which is more intensive during the times of low evapotranspiration. The presence of soil frost and snowfall events and the cycles of snow accumulation and melt strongly affect the winter and spring runoff behaviour, e.g., a long snow accumulation period increases the potential for snowmelt-induced large runoff volumes (Jamieson et al. 2003; Su et al. 2011; Turunen et al. 2015). High snow cover and increasing energy during late spring may generate rapid melt and high runoff peaks, whereas low snow cover and early melt may reduce spring runoff. Spring hydrological conditions can substantially impact environmental loads, since the occurrence of large runoff volumes on bare soil surface outside the growing season can cause elevated erosion and transport of sediment and nutrients to surface waters (e.g., Puustinen et al. 2007; Su et al. 2011; Øygarden et al. 2014). Future climate change in Finland and northern Europe may influence wintertime conditions, e.g., by increasing wintertime precipitation and freeze–thaw fluctuations, decreasing days with snow cover, and reducing snow depth (Jylhä et al. 2004, 2008). These changes are expected to affect runoff generation, erosion and nutrient losses during the off-season (Puustinen et al. 2007; Hägg et al. 2013; Øygarden et al. 2014; Huttunen et al. 2015). The changing hydrological conditions may challenge water protection policies, because more efficient means to control sediment and nutrient losses are needed outside the growing season (Huttunen et al. 2015; Rankinen et al. 2016). The hydrological processes leading to runoff generation are however complex, and future model predictions subject to uncertainties (Clark et al. 2016). Understanding runoff, erosion and nutrient loss processes at the source areas of environmental pollution are among the key factors for the assessment and development of mitigation measures.
It has been proposed that the weather conditions of the winter seasons are, to a large extent, controlled by the North Atlantic Oscillation (NAO) index, which describes the air pressure difference between the Icelandic region and the Azores region (e.g., Rodwell et al. 1999; Scaife et al. 2014). The NAO has two phases: the negative phase can be clearly associated with cold and snowy winters, and the positive phase with mild and wet winters in northern Europe (e.g., Kim & McCarl 2005). Climate change is projected to induce an upward trend to the NAO index, which increases the occurrence of positive NAO phases (Gillett et al. 2003). Thus, studying long-term hydrological data which include positive and negative NAO phases offers a method to assess how changing climate conditions can impact the hydrological processes of agricultural fields. However, such method has not been previously applied to quantify the impacts of changing climate on all water outflow components. Furthermore, the NAO phases may be predictable months ahead (Scaife et al. 2014), and hereby studying the hydrological processes and environmental loads under the different phases might provide the means to target the most effective water protection measures to years when the risks of high loads are elevated.
Process-based hydrological models provide an approach to quantify and understand cold region hydrological processes by providing a closure of the water balance and quantification of its components at field scale. The field-scale water balance results further support the assessment of nutrient losses and erosion, which are bound to hydrological variability. Models that support the simulation of snow and frozen soil processes are available for application in agricultural areas (e.g., Larsson & Jarvis 1999; Abrahamsen & Hansen 2000; Luo et al. 2000; Kroes et al. 2008; Warsta et al. 2012). One-dimensional (1D) models are well suited for describing the vertical transport of water stored in soil and snow, the main water balance components, and their temporal dynamics. However, 1D approaches are less suited for describing variable site topography, irregular drainage layouts, or assessing the impact of soil conservation measures which are distributed within the fields and, therefore, 2D or 3D modelling approaches have emerged (Mohanty et al. 1998; Gärdenäs et al. 2006; Hintikka et al. 2008). 3D hydrological models including wintertime process descriptions, such as freezing/thawing and snow accumulation/melt, and providing a description of spatially distributed features (e.g., Refsgaard et al. 2010; Šimůnek & Šejna 2011; Warsta et al. 2012) have the potential to enhance the understanding of runoff generation under varying meteorological conditions in high-latitude agricultural fields.
The objective of this study was to investigate the generation of drain discharge, surface runoff and groundwater flow in clayey agricultural fields during off-season periods with mild and cold winters, categorised based on the NAO index (NOAA-NWS 2015). The two studied field sections with different slopes (1% and 5%) are located in southern Finland, where the growing seasons are short and variable winter conditions with air temperature fluctuations around the freezing point are frequent, resulting in strongly altering snow accumulation and melt cycles. Water balance components of the fields were simulated based on recent modelling studies (Koivusalo et al. 2015; Turunen et al. 2015) that applied a spatially distributed 3D hydrological model to simulate water flow processes in the two field sections. The simulation results allowed the study of all water outflow components, including groundwater outflow, of which observations are rarely available. The available simulation results of the field sections covered a period of 7 years (2008–2014), including six off-season periods (September–April) which were the main focus of the analysis. The simulated snow variables, runoff components, as well as measured local meteorological conditions during off-seasons were analysed in relation to NAO.
MATERIALS AND METHODS
Site description and measurements
Location of the Gårdskulla Gård experimental field (a) and layout of two monitored field sections (b). Part (b) contains data from the National Land Survey of Finland (MML) Topographic Database 08/2015.
Location of the Gårdskulla Gård experimental field (a) and layout of two monitored field sections (b). Part (b) contains data from the National Land Survey of Finland (MML) Topographic Database 08/2015.
Turunen et al. (2015) studied the same two field sections and modelled the water balance of the fields during 2008–2012 using the hydrological FLUSH model (Warsta et al. 2013). This simulation period was further extended by 2 years by Koivusalo et al. (2015) who presented preliminary results of the 7-year simulation period. A short overview of the meteorological and the hydrological data used as model input and for the calibration–validation procedure is given here, while further details can be found in Turunen et al. (2015), Äijö et al. (2014) and Vakkilainen et al. (2008).
The subsurface drainage networks in the fields were instrumented to automatically measure drain discharge with a frequency of 15 min (Datawater WS vertical helix water meters, Maddalena, Povoletto, Italy). The area of the monitored subsurface drainage network was 5.7 ha in section 1 and 4.7 ha in section 2 (Figure 1(b)). In the downslope parts of the field sections, shallow interception drains (diameter 0.05 m) with coarse gravel as trench backfill were installed at the depth of 0.4 m to collect tillage layer runoff. The drains measuring tillage layer runoff (dark blue lines in Figure 1(b)) gathered both surface runoff and shallow lateral water seepage at the top soil layer (tillage layer). The tillage layer runoff was recorded with a 15 min time interval and the local drainage area of the drains was 3.3 ha in section 1 and 3.0 ha in section 2. The monitored field drainage areas (for subsurface drain discharge and tillage layer runoff) in sections 1 and 2 were delineated on the basis of the instrumented subsurface drain networks and the terrain topography controlling the surface runoff and tillage layer flow (Äijö et al. 2014).
In addition to the drain discharge and tillage layer runoff measurements in the two sections, precipitation, snow water equivalent (SWE) and the level of the groundwater table were manually measured on site. Soil characteristics (porosity, water retention curves) were determined from soil samples. Rainfall was recorded every 15 min using a RAINEW 111 tipping bucket rain gauge (RainWise Inc., Bar Harbor, ME, USA), while winter precipitation (including snowfall) was measured manually weekly/biweekly. SWE and groundwater table level were manually observed weekly/biweekly during field visits from 2008 to 2014. Meteorological variables, including year-round precipitation (used to disaggregate manual winter precipitation measurements to hourly data), air temperature, relative humidity, wind speed and global radiation were obtained from three weather stations of the Finnish Meteorological Institute, located at the distances of 10–47 km (Turunen et al. 2015). The form of precipitation (snowfall or rain) was described as a function of air temperature, and rainfall and snowfall estimates were corrected for gauging errors with coefficients of 1.05 and 1.3, respectively (Førland et al. 1996).
Field hydrology simulations
The simulated hourly values for water balance components were available from the two field sections for the 7-year study period, 2008–2014 (Koivusalo et al. 2015; Turunen et al. 2015). The simulations were conducted with FLUSH, which is a spatially distributed hydrological model developed for simulating water flow processes in clayey subsurface drained agricultural fields (Warsta 2011; Warsta et al. 2013). FLUSH divides the computational area into 2D overland and 3D subsurface domains. Water flow in the subsurface domain follows the dual permeability approach, in which the total pore space is divided into mobile soil matrix and macropore systems. The model takes into account the dynamic changes of soil macroporosity by simulating the soil shrinking and swelling processes (Kroes et al. 2008) caused by drying and wetting of clay soil. The model simulates wintertime processes based on an energy balance snow model (Koivusalo et al. 2001) and a frozen soil description (Karvonen 1988). FLUSH produces model outputs including hourly runoff components (subsurface drain discharge, surface runoff, groundwater outflow), soil moisture conditions (e.g., groundwater depth), snow variables (e.g., snowfall, SWE) and soil temperatures.
The field areas used for the simulations (Koivusalo et al. 2015; Turunen et al. 2015) are shown in Figure 1(b). In section 1, the area comprises 12.4 ha including the upslope area outside of the field, as Turunen et al. (2015) found that this area has a hydrological connection to the field. The field receives groundwater flow from the upslope area (4.6 ha), but surface runoff from the upslope area is cut off by a roadside open ditch. The area of the steeper field section 2 is 7.8 ha. The boundary conditions of the model domains are described by no-flow interfaces at the upslope ends of the simulated areas and the Kirkkojoki stream water level at the downslope ends (for a more complete description see Turunen et al. (2015)).
Measured and simulated annual cumulative drain discharge and tillage layer runoff in field sections 1 with 1% slope (a) and 2 with 5% slope (b) during 2008–2014 showing calibration and validation periods, as well as land use type. The modified efficiency of Legates & McCabe (1999) is shown for the drain discharge and tillage layer runoff for every year. The fluxes are computed as flow volume per monitored drainage areas.
Measured and simulated annual cumulative drain discharge and tillage layer runoff in field sections 1 with 1% slope (a) and 2 with 5% slope (b) during 2008–2014 showing calibration and validation periods, as well as land use type. The modified efficiency of Legates & McCabe (1999) is shown for the drain discharge and tillage layer runoff for every year. The fluxes are computed as flow volume per monitored drainage areas.
Analysis of NAO, meteorology and field water balance
The NAO index (NOAA-NWS 2015) was used to describe the six off-season periods (September–April) of the 7-year study period. The mean NAO indices for the off-season periods were computed from the monthly NAO indices available from NOAA-NWS (2015). A positive NAO index was associated with mild winters and a negative NAO index with cold winters.
The off-season NAO index was evaluated against mean off-season temperature, precipitation and snow cover variables (fraction of snowfall, mean SWE) to understand the effect of NAO on local climate. Temperature and precipitation were measured variables while snow cover variables were available from FLUSH outputs. The relation of NAO to these variables was explored by means of linear correlation (R).
Runoff generation during the off-seasons in the two field sections was examined in relation to NAO in order to identify differences between mild and cold winters. Daily runoff components aggregated from the hourly model outputs produced by FLUSH were used for this purpose. In this approach, the use of continuous hydrological variables that represented unchanged land use enabled an extended analysis beyond the monitored subareas of the fields, and included runoff components (groundwater outflow) that were not measured in the field. The analysed time series included SWE, subsurface drain discharge, tillage layer runoff and subsurface groundwater flow to the Kirkkojoki River (see Figure 1). In this analysis, extending over the entire fields, the tillage layer runoff included surface runoff (to ditches and stream) and water seepage to the interception drains and the shallow open ditches (see Figure 1), whereas groundwater flow represented the seepage of water into the Kirkkojoki stream. Runoff components were analysed as mean monthly outflows of off-season periods against NAO and as daily values categorising them based on positive and negative off-season NAO indices.
RESULTS AND DISCUSSION
Overview of field water balance
Cumulative water balance components from 2008 to 2014 in Gårdskulla Gård for the flat field section 1 including the upslope area ((a) slope 1%, 14.6 ha) and the steep field section 2 ((b) slope 5%, 7.8 ha).
Cumulative water balance components from 2008 to 2014 in Gårdskulla Gård for the flat field section 1 including the upslope area ((a) slope 1%, 14.6 ha) and the steep field section 2 ((b) slope 5%, 7.8 ha).
NAO against mean off-season variables
Mean monthly air temperature (a), precipitation (b), fraction of snowfall and mean SWE (c), modelled groundwater outflow (d), modelled subsurface drain discharge (e), and modelled tillage layer runoff (f) in the two field sections (1% and 5% slope) of Gårdskulla Gård during September–April as a function of mean monthly off-season NAO index for the years 2008–2014.
Mean monthly air temperature (a), precipitation (b), fraction of snowfall and mean SWE (c), modelled groundwater outflow (d), modelled subsurface drain discharge (e), and modelled tillage layer runoff (f) in the two field sections (1% and 5% slope) of Gårdskulla Gård during September–April as a function of mean monthly off-season NAO index for the years 2008–2014.
Figure 4(d)–4(f) show the mean monthly off-season runoff components from the two sections with different slopes. Here, the runoff components were computed by dividing the flow volumes by the area of the cultivated field (7.8 ha in both field sections). Precipitation was clearly the main driver of the runoff components, which all had a significant correlation with precipitation (p < 0.04). The difference between the sections is seen as higher drain discharge and tillage layer runoff from the flat field section 1 and higher groundwater flow from the steep field section 2. The groundwater flow entering section 1 from the upslope area increases subsurface drain discharge in section 1. The differences between the mild and cold periods cannot be clearly seen in runoff components that are aggregated measures of the entire off-season periods (Figure 4(d)–4(f)). However, some of the runoff components in the steeper field section 2 (Figure 4(d) and 4(e)) showed a stronger correlation with NAO than precipitation, which may be a reflection of air temperature impact on the form of precipitation. The clearly warmer off-seasons and weakly higher precipitation in Gårdskulla during positive NAO was in line with Kim & McCarl (2005), who reported that the relation between NAO and off-season weather patterns reflects the warmer and wetter European winters during positive NAO.
Daily SWE and runoff during mild and cold off-seasons
Envelopes of SWE (a) and total runoff from the field sections with 1% slope (b) and 5% slope (c) during periods (September–April) of positive and negative NAO in the Gårdskulla Gård experimental field during 2008–2014.
Envelopes of SWE (a) and total runoff from the field sections with 1% slope (b) and 5% slope (c) during periods (September–April) of positive and negative NAO in the Gårdskulla Gård experimental field during 2008–2014.
Upper 50% cumulative distribution of daily subsurface drain discharge intensities during the periods with positive and negative NAO index for the section with 1% slope (a) and 5% slope (d), cumulative distribution of daily groundwater outflow intensities for the section with 1% slope (b) and 5% slope (e), and cumulative distribution of daily tillage layer runoff intensities for the section with 1% slope (c) and 5% slope (f) in the Gårdskulla Gård experimental field during 2008–2014.
Upper 50% cumulative distribution of daily subsurface drain discharge intensities during the periods with positive and negative NAO index for the section with 1% slope (a) and 5% slope (d), cumulative distribution of daily groundwater outflow intensities for the section with 1% slope (b) and 5% slope (e), and cumulative distribution of daily tillage layer runoff intensities for the section with 1% slope (c) and 5% slope (f) in the Gårdskulla Gård experimental field during 2008–2014.
Compared to subsurface drain discharge, the distributions of daily groundwater outflow from the field sections (Figure 6(b) and 6(e)) showed a similar relation to weather conditions. The slope of the field section and the upslope hydrological connection in section 1 had a stronger impact on the groundwater outflow volume than the changing NAO and snow conditions. This demonstrated how the field or catchment characteristics, such as the terrain topography, can impact how the hydrology of the area responds to the NAO phases and snow conditions. The slow hydrological processes of those areas where groundwater outflow dominates may be less sensitive to the changing NAO phases than in other types of areas. The distribution of daily tillage layer runoff showed how increasing slope decreased its peak values. Distributions of tillage layer runoff (Figure 6(c) and 6(f)) showed no clear differences between periods of mild and cold winters.
The results of this study indicate the high sensitivity of snow conditions, water balance and runoff generation to the variation in temperature regime at high latitude regions. Winter air temperature fluctuates around the freezing point affecting the snow conditions and, as a result, the water balance and runoff generation become sensitive to variation in temperature regime. These winters are crucial periods for water protection measures, as there have been concerns about increasing sediment and nutrient losses during mild winters (e.g., Granlund et al. 2007; Øygarden et al. 2014). For example, Rankinen et al. (2016) found that increasing total nitrogen concentration in large river basins could be explained by climatic factors and Puustinen et al. (2007) reported higher sediment loads during mild rather than cold winters in agricultural fields in Finland. Øygarden et al. (2014) summarised catchment studies and climate change scenarios from the Baltic region. They showed nitrogen fluxes correlated with runoff volumes, even though the variation between different catchments was large. In the light of the current results, the climate warming and increasing winter precipitation pose the highest risks on the edge of snow-covered areas, such as southern coastal Finland, where frequent occurrence of runoff events may become common.
CONCLUSIONS
FLUSH was applied to produce a quantification of water balance in two clayey agricultural field sections with different slopes in Gårdskulla Gård in southern Finland, where the growing seasons are short and variable winter conditions with air temperature fluctuations around the freezing point are frequent. The 7-year study period included six off-season periods (September–April), of which three were characterised by the positive mean NAO index (mild winter) and three by the negative mean NAO index (cold winter). Evapotranspiration in the studied high-latitude site was the dominant water component during the growing seasons from May to August, while drain discharge dominated the water balance during the periods outside the growing season from September to April.
Annual differences in precipitation were strongly reflected in the annual volumes of drain discharge, whereas annual evapotranspiration showed lower variation between the years. The increase of slope in the field area decreased the volume of subsurface drain discharge and increased groundwater outflow to the stream at the downslope end of the field. The variability in the form of precipitation and further in snow accumulation was related to the weather patterns classified by the NAO index during the period outside of the growing season.
Mild periods with positive NAO were associated with the increased frequency of off-season runoff events, which were sustained throughout wet periods with minor snow depths. Tillage layer runoff in the near surface layers of the field showed higher intensities on the flat field section, but its occurrence depended more on the intensities of rainfall and snowmelt instead of the periodical weather patterns. Understanding and quantifying the water balance and all major water outflow components through periods of different weather patterns is essential in choosing suitable mitigation measures and control nutrient and sediment losses. The NAO index as a measure of mild (positive NAO) and cold (negative NAO) periods demonstrated how the snow accumulation and runoff components respond to alternating wet and cold winter conditions.
ACKNOWLEDGEMENTS
The study was funded by the Aalto University School of Engineering, the Finnish Drainage Research Foundation, the Ministry of Agriculture and Forestry, Maa-ja vesitekniikan tuki ry, Sven Hallin Research Foundation, and Academy of Finland. We acknowledge CSC – IT Center for Science Ltd for the allocation of computational resources.