Title : Impact of capillary rise and recirculation on simulated crop yields

Upward soil water flow is a vital supply of water to crops. The purpose of this study was to 10 determine if upward flow and recirculated percolation water can be quantified separately, and to determine the contribution of capillary rise and recirculated water to crop yield and groundwater recharge. Therefore we performed impact analyses of various soil water flow regimes on grass, maize and potato yields in the Dutch delta. Flow regimes are characterised by soil composition and groundwater depth and derived from a national soil database. The 15 intermittent occurrence of upward flow and its influence on crop growth are simulated with the combined SWAP-WOFOST model using various boundary conditions. Case studies and model experiments are used to illustrate impact of upward flow on yield and crop growth. This impact is clearly present in situations with relatively shallow groundwater levels (85% of the Netherlands), where capillary rise is a well-known source of upward flow; but also in free20 draining situations the impact of upward flow is considerable. In the latter case recirculated percolation water is the flow source. To make this impact explicit we implemented a synthetic modelling option that stops upward flow from reaching the root zone, without inhibiting percolation. Such a hypothetically moisture-stressed situation compared to a natural one in the presence of shallow groundwater shows mean yield reductions for grassland, maize and 25 potatoes of respectively 26, 3 and 14 % or respectively about 3.7, 0.3 and 1.5 ton dry matter per ha. About half of the withheld water behind these yield effects comes from recirculated percolation water as occurs in free drainage conditions and the other half comes from increased upward capillary rise. Soil water and crop growth modelling should consider both capillary rise from groundwater and recirculation of percolation water as this improves the 30 accuracy of yield simulations. This also improves the accuracy of the simulated groundwater recharge: neglecting these processes causes overestimates of 17% for grassland and 46% for potatoes, or 63 and 34 mm year, respectively.

determine if upward flow and recirculated percolation water can be quantified separately, and to determine the contribution of capillary rise and recirculated water to crop yield and groundwater recharge. Therefore we performed impact analyses of various soil water flow regimes on grass, maize and potato yields in the Dutch delta. Flow regimes are characterised by soil composition and groundwater depth and derived from a national soil database. The 15 intermittent occurrence of upward flow and its influence on crop growth are simulated with the combined SWAP-WOFOST model using various boundary conditions. Case studies and model experiments are used to illustrate impact of upward flow on yield and crop growth. This impact is clearly present in situations with relatively shallow groundwater levels (85% of the Netherlands), where capillary rise is a well-known source of upward flow; but also in free-20 draining situations the impact of upward flow is considerable. In the latter case recirculated percolation water is the flow source. To make this impact explicit we implemented a synthetic modelling option that stops upward flow from reaching the root zone, without inhibiting percolation. Such a hypothetically moisture-stressed situation compared to a natural one in the presence of shallow groundwater shows mean yield reductions for grassland, maize and 25 potatoes of respectively 26, 3 and 14 % or respectively about 3.7, 0.3 and 1.5 ton dry matter per ha. About half of the withheld water behind these yield effects comes from recirculated percolation water as occurs in free drainage conditions and the other half comes from increased upward capillary rise. Soil water and crop growth modelling should consider both capillary rise from groundwater and recirculation of percolation water as this improves the 30 accuracy of yield simulations. This also improves the accuracy of the simulated groundwater recharge: neglecting these processes causes overestimates of 17% for grassland and 46% for potatoes, or 63 and 34 mm year -1 , respectively.

Introduction
Crop growth strongly depends on soil moisture conditions. Climate variables determine these conditions through rain that penetrates directly into the root zone or comes available via lateral flow. The moisture distribution in the soil strongly depends on soil physical properties that 40 determine vertical flow. Upward soil water flow becomes an especially vital supply term of a crop when the soil water potential gradient induced by the root-extraction manages to bridge the distance to the capillary fringe, inducing increased capillary rise. In this paper we follow the definition of capillary rise, given by SSSA (2008), as the "phenomenon that occurs when small pores which reduce the water potential are in contact with free water". This implies that 45 capillary rise as a source for upward flow to crop roots requires the presence of a groundwater table. In conditions without a groundwater table there may also be a contribution of upward flow to crop roots through the process of recirculation. Recirculation is a known process discussed already by Feodoroff (Rijtema and Wassink, 1969) but has never been quantified.
We quantified recirculation separately from capillary rise using model experiments.

50
The contribution of (intermittent) upward flow to the total water budget can be significant. For example Kowalik (2006) mentions that during the grass growing season, in soils with the groundwater close to the soil surface (Aquepts) the capillary rise induced by root extraction varies between 60 and 150 mm per year. Babajimopoulos et al. (2007) found that under the specific field conditions about 3.6 mm/day of the water in the root zone originated from the 55 shallow water table, which amounts to about 18% of the water transpired by a maize crop. Fan et al. (2013) analysed the groundwater depth globally and concluded that shallow groundwater influences 22 to 32% of global land area, and that 7 to 17% of this area has a water table within or close to plant rooting depths, suggesting a widespread influence of groundwater on crops. This is especially the case in delta areas where high population 60 densities occur and agriculture is the predominant land use. Wu et al. (2015) showed that capillary rise plays a main role in supplying the vegetation throughout the season with water, hence a strong dependence of vegetation upon groundwater. Han et al. (2015) (Videla Mensegue et al., 2015).
In 85% of the area in the Netherlands the average groundwater table is less than 2 meter below the soil surface in (De Vries, 2007), where root extraction can induce capillary rise from 75 groundwater. Wesseling and Feddes (2006) report that in summers with a high evapotranspiration demand, crops partially depend on water supply from soil profile storage and induced capillary rise. Van der Gaast et al. (2009), applying the method of Wesseling (1991), found for the Netherlands a maximum capillary flow of 2 mm/d to the root zone in loamy soils where the groundwater level is at 2.5 meter below the soil surface.

80
Although the contribution of capillary rise to the total water budget can be significant, it is an often neglected part of the crop water demand in situations of shallow groundwater levels (Awan et al., 2014). The capillary properties of a soil strongly depend on soil type. Rijtema (1971) estimated that loamy soils have an almost 2 times higher capillary rise than sandy soils.

85
Integrated approaches are needed to relate water availability to crop yield prognosis (Van der Ploeg and Teuling, 2013;Norman, 2013). The importance of capillary rise as supplier of water to crops has been shown by many researchers (e.g. Hooghoudt, 1937;Huo et al., 2012;Talebnejad, and Sepaskhah, 2015;Han et al., 2015); however we found only a few studies 90 that use an integrated modelling approach (Xu et al., 2013;Zipper et al. 2015) to quantify capillary rise for different hydrological conditions (including free drainage) using physically based approaches. In this study we explicitly consider the effect of crop type, soil type, weather year and drainage condition on capillary rise. Zipper et al. (2015) introduced the concept of groundwater yield subsidy as the increase in harvested yield (kg/ha -1 ) in the 95 presence of shallow groundwater compared to free drainage conditions. Following their line we introduce the concept of soil moisture yield subsidy as additional yield increase in free drainage conditions due to recirculation of percolated soil moisture.
The driving force for induced capillary rise and recirculation is the difference in soil water 100 potential, referred to as heads, at different soil depths. There are several models available that solve these head differences numerically. Ahuja et al. (2014) evaluated 11 models commonly applied for agricultural water management. Six of these models use simple 'bucket' approaches for water storage and have in some cases been extended with more or less empirical options for capillary rise. Five models have the ability to numerically solve Richards 105 equation for water movement in the soil. Examples are HYDRUS (Šimůnek et al., 2008) and SWAP (Feddes et al., 1988. We applied the integrated model SWAP-WOFOST (acronyms for Soil Water Atmosphere Plant -WOrld FOod Studies) to solve head differences and crop yield simulations. Kroes and Supit (2011) applied the same integrated model to quantify the impact of increased 110 groundwater salinity on drought and oxygen of grassland yields in the Netherlands. They recommended further analyses using different crops and different boundary conditions. We now apply this model with different boundary conditions using 45 years of observed weather and three different crops. For the lower boundary we use different hydrologic conditions that influence the vertical flow. For the soil system itself we use a wide range of soil physical 115 conditions. The importance of the soil system was already stated by several authors like Supit (2000). We build on their suggestions and apply the tools for different crops and boundary conditions. Before we applied the model to different boundary conditions we validated it at field scale.

120
This paper quantifies the effects of (intermittent) upward flow on crop growth under different conditions of soil hydrology, soil type and weather. The effects are separately quantified in terms of flow source, namely capillary rise and recirculated percolation water. Therefore we introduced a synthetic model option and performed a numerical experiment. We studied forage maize, grassland and potatoes and we hypothesize that neglecting upward flow will 125 result in neglecting a considerable amount of soil moisture that is available for crop growth.
We quantify this amount and show the importance of including upward flow for crop growth modelling. Our main research questions are: i) Can upward flow with capillary rise and recirculated percolation water as source be quantified separately?, ii) What is the contribution of capillary rise and recirculated water to crop yield and groundwater recharge? We applied the coupled SWAP and WOFOST modeling system, using a one day time step.
SWAP Kroes et al., 2017) is a one-dimensional physically based transport model for water, heat and solute in the saturated and unsaturated zone, and includes modules for simulating irrigation practices. The first version of SWAP, called SWATRE, was developed by Feddes et al. (1978). This version also included a module for crop production,

140
CROPR that applied principles of C.T. de Wit (1965) and is still applied in several countries. SWAP simulates the unsaturated and saturated water flow in the upper part of the soil system, using a numerical solution of the Richards equation: where:  is volumetric water content (cm 3 cm -3 ), t is time (d), K(h) is hydraulic conductivity (cm

150
The numerical solution of this equation uses variable time steps that depend on boundary conditions and an iteration scheme. For example, high fluxes require time steps that are much smaller than 1 day (see Kroes et al, 2017 for a detailed explanation).
Root water extraction and lateral exchange with surface water were accounted for. In this 155 study we did not use the option to exchange water flow with macro pores.
The soil hydraulics were described by the Mualem-van Genuchten relations and the potential evapotranspiration was calculated with the Penman-Monteith equation (Allen et al., 1998). At the bottom boundary water fluxes, supplied by a separate regional hydrological model were used. Drainage and infiltration through the lateral boundary accounted for the flow to surface 160 water. The surface water system was simulated using a simplified, weir controlled, water balance. Note that the surface water system in its turn interacted with the groundwater system.
In previous years, SWAP has been successfully used to study soil-water-atmosphere-plant relationships in many locations with various boundary conditions (e.g. Feddes et al., 1988;Bastiaanssen et al., 2007). See Van Dam et al. (2008) for an overview. A recent list is available applied in many studies (e.g. Rötter, 1993;Van Ittersum et al., 2003;de Wit and Van Diepen, 2008;Supit et al., 2012;De Wit et al., 2012). Crop assimilation was calculated as function of solar radiation and temperature, using a 3 point Gaussian integration method accounting for leaf angle distribution and extinction of direct and diffuse light. The assimilation was reduced when water stress occurred. Subsequently, the maintenance respiration was subtracted and 175 the remaining assimilates were partitioned over the plant organs (i.e. leaves, stems, roots and storage organs). For maize and potatoes the partitioning was development stage dependent.
For perennial grass however, a constant partitioning factor was assumed. By integrating the difference between growth and senescence rates over time, dry weights of various plant organs were established.

180
In SWAP-WOFOST, crop assimilation depends on the ambient CO2 concentration as well (see: Kroes and Supit, 2011;Supit et al., 2012). To account for unknown residual stress caused by diseases, pests and/or weeds an additional assimilation reduction factor was introduced. The rooting density decreased exponentially with depth. To withdraw water from deeper soil layers for crop uptake a form of compensatory root uptake was used in case the upper part of the soil was very dry (Jarvis, 2011). The increasing atmospheric CO2 concentrations during relatively long historical simulation periods (>20 years) was accounted for.

Case studies for validation
190 SWAP-WOFOST was validated using results of 7 case studies at 6 locations in the Netherlands ( Figure 1)  205 the dry part of the reduction function proposed by Feddes et al. (1978). Drought stress was absent when the soil pressure head h exceeded the critical value of h3. Drought stress increased linearly between h3 and at h4 (wilting point). The critical pressure head h3 differed between lower and higher potential transpiration (respectively h3l and h3h) rates. In conditions with drought or oxygen stress, the reduction in stressed parts was partly compensated by 210 extra root water uptake in those parts of the root zone with more favorable soil moisture conditions (Jarvis, 1989).
For all cases a so-called management factor was used to close the gap between observed and actual yield. The input crop parameters for maize only differed with respect to the management factor which ranges from 0.85-0.95. The management factors were relatively 215 high because the case study locations have good management. It is very likely that we missed some processes even though our modelling approach is mechanistic, because it is still relatively simple. Some processes like pests and diseases were not included and may have played a role in the field; the calibration was done on experimental farms where the impact from diseases and pests was minimal.

220
For potatoes the input crop parameters were kept the same for all 3 cases (Table 1). Maximum rooting depth for grassland, maize and potatoes were respectively 40, 100 and 50 cm.
Soil water conditions were different for all locations and boundary conditions varied, depending on local situation and available data (Table 1). In most cases a Cauchy bottom boundary condition was applied using a hydraulic head based on piezometer observations 225 from the Dutch Geological Survey (https://www.dinoloket.nl/). Observed groundwater levels were used as lower boundary condition for Borgerswold (crop: potato). In 2 cases a lateral boundary condition was applied with drainage to a surface water system (Table 1). The simulation results were analysed using an R-package (Bigiarini, 2013) and the statistics are presented in Table 2.

Case studies for validation
The first 2 case studies are from one location (De Marke) where a grassland-maize rotation was practised. The results show that the hydrological conditions ( Figure 4 and Table 2) were 300 simulated accurately for those years for which observed data were available (1991)(1992)(1993)(1994)(1995).
From 1995-1997 the groundwater levels dropped as a result of low precipitation (about 700 mm/year). The fall of the year 1998 showed rising groundwater levels that corresponded well with very wet conditions at that moment. The simulated grassland yields were overestimated by 133 kg.ha -1 DM and the simulated maize yields were underestimated by 257 kg.ha -1 DM 305 which differences were well within acceptable ranges ( Figure 5 and Table 2).
For the other 2 maize case studies (C-Maize and D-Maize) groundwater levels and soil moisture were well simulated ( Table 2). The simulated maize yields (Table 2) were less acceptable for case C-Maize as indicated by a zero or negative Nash-Sutcliffe efficiency (NS) which suggests that the observed mean was a better predictor than the model. One should 310 consider that the NS efficiency is sensitive to sample size and outliers. In 1976, a very dry year, the soil hydrology dynamics and the resulting yield were well captured. The yield of case study D-Maize had a small bias of 333 kg.ha -1 DM between observed and simulated.
The simulated hydrological conditions for the 3 fields of the potato-cases R-Potato and V-Potato showed a good fit with the observed ( Table 2). The simulated yields (Table 2)  However one has to bear in mind that perfect calibration is not the objective of this study. We used calibration values from earlier studies (Kroes et al., 2015 andHack et al., 2016). No detailed assimilation measurements were executed on the fields and the meteorological data was not measured on site, but taken from meteorological stations sometimes more than 30 325 km away. Furthermore, no detailed information concerning fertilizer applications and soil carbon weres available, therefore we considered it constant in time.
Even though some yields were not accurate enough to satisfy statistical criteria for good model performance, we think that the dynamics of soil hydrology and crop yield were acceptably captured. With more field information and calibration a better result could be achieved but we 330 think that current tuning of SWAP-WOFOST for the 3 crops allowed an application at a larger scale with various hydrological boundary conditions.
Before the analysis at a larger scale we simulated the impact of upward flow for the case studies. We carried out additional simulations without upward flow towards the root zone,

335
using the specially programmed synthetic model option. Results of these 3 cases are given in Table 3 for the situation with and without upward flow. This table shows that suppressing upward flow lowered yields by 6, 3 and 20% respectively for grassland, maize and potato .
The groundwater recharge was reduced with respectively 3, 4 and 94% (Table 3). Detailed results can be found in the supplementary material (S1). In supplementary material S4 input 340 data for case 3 (V-Potato) can be found. In a next step we carried out a larger scale experiment to quantify this impact for different soil crop and climate conditions.

Soil crop experiment to analyse the role of capillary rise
The 3 crops from the case studies were simulated with 72 soils from the national database  Table 4). In free-draining soils the variation of upward flow to the root zone ranged from about 10 mm in wet and cold to 120 mm in dry and warm years with a high evaporative demand (Figures 6, upper part). In general upward flow was highest in loamy soils where soil physical conditions were optimal. Especially in the presence of a groundwater level differences in upward flow between soils were relatively small compared to differences 360 among years and within one grouped soil type (Figure 7, lower part).
The upward flow was inversely related to the rooting depth: the larger the rooting depth, the smaller the upward flow. Grassland, potatoes and maize had rooting depths of respectively 40, 50 and 100 cm and an upward flow of respectively 194, 112 and 74 mm per growth season (Table 4). Note that the high value for perennial grassland was also caused by a much longer 365 growing season. The percolation was highest for grassland for the same reasons (Table 4).
These high values were largely due to the precipitation excess during winter in the Netherlands.
Upward seepage across the bottom boundary did not occur in the free-drainage conditions 370 ( Figure 2 a and b). Leaching was highest (Table 4) in the synthetic free-drainage condition without capillary rise (Figure 2 a). Note that the values in Table 4 for seepage and leaching were given for a calendar year whereas the other mean values were given for a growing season. Yearly values were used for the bottom boundary because these values give an indication for the yearly deeper groundwater recharge which may also be influenced by variations of vertical fluxes close to the root zone during the remainder of the year. The leaching flux at 5.5 m depth (Table 4, qleaching) increased when upward flow was suppressed (lower transpiration, more groundwater recharge), with respectively 44, 2 and 16 mm.year -1 for grassland, maize and potatoes. The shallow groundwater in Dutch conditions (Figure 2 c) often does not have leaching at greater depth because excess precipitation or upward 380 seepage is discharged via drainage systems. The average condition we used had no leaching but seepage of 227, 155 and 291 mm.year -1 for grassland, maize and potatoes (Table 4, qseepage).
As can be expected, the synthetic condition without upward flow and without groundwater 385 (Figure 2 a), had the lowest simulated mean yields for all crops ( Table 4). The highest mean yields were simulated when average groundwater situations including capillary rise were considered (Table 4, Ave). The relative mean yield increase was lowest for maize and highest for grassland (Table 5) which was probably caused by the difference in rooting depth.

390
The simulation results with 3 different lower boundary conditions (Figure 2 conditions a, b and c) were also compared by subtraction. The subtraction enables a quantification of the contribution of the 2 different sources of upward flow: groundwater and recirculating percolation water.
The elimination of recirculating percolation water to the root zone in free drainage conditions 395 (synthetic condition a compared to b, Figure 2) reduced grassland, maize and potato yields with respectively 14, 0 and 7 % ( Table 5). The higher yields were caused by upward flow using recirculating percolation water as source.
A comparison between situations with free drainage (condition b, Figure 2) with average groundwater levels (condition c, Figure 2) showed a similar yield reduction: respectively 14, 2 400 and 8 %. The higher yields were caused by capillary rise with groundwater and recirculation as source.
When one compares situations with free-drainage conditions without upward flow (synthetic condition a, Figure 2) with average groundwater levels (condition c) yield-reductions of grassland, maize and potatoes were respectively 26, 3 and 14 % (Table 5) or respectively 405 about 3.7, 0.3 and 1.5 ton.ha -1 dry matter (Table 4). These yield differences quantify the contribution of the sum of the two different sources of upward flow: groundwater and recirculating percolation water.
The impact of upward flow on groundwater recharge was highest for potatoes and lowest for maize. For grassland, maize and potatoes differences between downward flux across the 410 bottom of the root zone (qpercolation in Figure 2) of 3 hydrological conditions were calculated of respectively 17, -11 and 46 % (qpercolation in Table 5) or 63, -5 and 34 mm (qpercolation in Table 4).
Low recharge values for maize were caused by deeper rooting systems which reduced these differences because groundwater levels were closer to the bottom of the root zone. For potatoes this difference in yield did reach values of more than 4 ton.ha -1 dry matter in stress 415 conditions (Table 6). The results are presented in more detail in the supplementary material (S3).

420
The case studies and soil-crop experiments in this paper demonstrate the combined interaction of recirculation and capillary rise on crop yields. This impact is clearly present in situations where a groundwater level is present (85% of NL) but also in free-draining situations the impact of upward flow is considerable. According to our simulation results, grassland, 425 maize and potato yields increased with respectively 14, 0 and 7% in free drainage conditions when upward flow wass included (Table 5) Crop models that apply tipping bucket approach consider the soil system as a reservoir with only percolation and no upward flow (an overview with a model comparison is provided by Ahuja et al, 2014). Such models do not account for soil moisture redistribution within and 440 below the root zone. Similar to Guderle and Hildebrand (2015) our simulation results show that a detailed vertical flow improves predictions of root water uptake. Tipping bucket models generally overestimate drought stress and groundwater recharge and subsequently underestimate crop yield. The irrigation demand may be overestimated as well. The high percolation may also result in overestimation of groundwater recharge (leaching).

445
Groundwater depth is important, because it determines the distance that the capillary flux has to bridge to reach the root zone and should be accounted for in crop modelling.
In the ideal situation one should compare the bucket approach to the approach with full simulation of capillary rise and recirculation using independent data sets. However the measured data sets were insufficient to calibrate and validate the soil and crop parameters in such detail that they allow proper statistical evaluation of the two approaches. The calibration of both model approaches had too much freedom with the available datasets, which upset a reliable validation. Therefore we used the measured data sets to illustrate that with common soil and crop input values SWAP-WOFOST yielded realistic and plausible results for the crops considered in this study. Further, crop growth and soil water flow were simulated by SWAP-

455
WOFOST with state of the art concepts. Therefore we may expect that the model itself can be used to show the effect on crop yield of different boundary conditions with respect to zero flux, recirculation and capillary rise.
Our hypothesis is that the process of recirculation makes crop modelling more accurate.
To demonstrate and support our hypothesis we added another case study. This is reported in 460 section S5 of the supplementary material. In this section we demonstrated the difference in soil water pressure head in the upper part of the root zone as caused by drying of the soil due to a lack of recirculating water in the hydrological condition ( Figure 2a). This resulted in a lowering of average yields with 609 kg/ha (from 7132 to 7741 kg/ha DM, which is about 9% yield reduction due to recirculation. This supports the recommendation to use tools that support this process of recirculation in conditions where the vertical water fluxes across the root zone is relatively high. This will clearly be the case in delta regions where you have occasionally a precipitation excess. A bucket approach generally underestimates water availability in the rooting zone and 470 consequently overestimates drought stress (Boogaard et al., 2013). We suggest to generate additional relations about the contribution of recirculation and capillary rise to upward flow to the root zone. Such an approach has been used in AQUACROP to derive a relation between capillary rise and groundwater (Van Gaelen et al., 2017). Another approach is to calibrate the conceptual parameters of a bucket model with agro-hydrological models like SWAP as done 475 by Romano et al. (2011).
Our analysis shows that soil properties and soil profile layering are important because differences in soil hydraulic properties influence vertical water flow. High upward flow values were found in loamy soils as was expected (Table 6, max row), but if water stress was high and upward flow was low the influence of soil type decreased and low upward flow values 480 were found for loamy soils (Table 6, min row). Comparing the minimum yield values it showed that there was a large difference between these soil types in free-drainage conditions with and without upward flow. This means that the storage capacity of loamy soils was larger than the one of sandy soils as could be expected. The yield variation between soil types in water stress conditions was large and illustrated the need for a proper soil schematization especially in 485 stress full hydrological conditions. As the influence of recirculation increased, the yield variation became less and the influence of soil type decreased. In situations without water stress the soil type was less important. In conditions where groundwater and capillary rise occurred (Ave) yield variation was hardly influenced by soil type.
Therefore modelling concepts should consider dynamic interactions between soil water and 490 crop growth. Crop models in general should consider recirculation of soil water and, especially in low lying regions like deltas, groundwater dynamics should be considered as well.
Precipitation, soil texture and water table depth jointly affected the amount of groundwater recharge and time-lag between water input and groundwater recharge (Ma et al., 2015). We 495 quantified some of these issues, but several items remain, such as the impact of rooting depth on crop yield and transpiration. Also soil and water management practises like ploughing and irrigation, were not considered. Furthermore the rooting pattern needed a more detailed analysis; we applied an exponential decrease of root density and compensation of root uptake according to Jarvis (2011) but the macroscopic root water uptake concept was still simple and 500 may require a more detailed analyses (Dos Santos et al. 2017). Another item we neglected is the preferential flow of water by the occurrence of non-capillary sized macropores (Bouma, 1961, Feddes, 1988, which is relevant in especially clay soils. Hysteresis of the water retention function was also not considered. An additional analysis of these issues is recommended, especially the impact of different rooting patterns on capillary rise should be 505 addressed.
The impact of soil type on yield increased when environmental conditions became dryer; situations without groundwater and without recirculation had less yield and higher yield variation than situations where groundwater influenced capillary rise (For detailed information 510 on results see the supplementary material S1 and S3).

515
We quantified the impact of upward flow on crop yields of grassland, maize and potatoes in layered soils. We compared situations with average groundwater levels with free-drainage conditions with and without upward flow. The largest impact of upward flow on crop yields was found when one compares situations with average groundwater levels with free drainage conditions without upward flow. From these differences one may conclude that neglecting 520 upward flow has a large impact on simulated yields and water balance calculations especially in regions where shallow groundwater occurs. The comparison showed long term average yield-reductions of grassland, maize and potatoes of respectively 26, 3 and 14 % (Table 5) or respectively 3.7, 0.3 and 1.5 ton Dry Matter per ha (Table 4). Reduction of the percolation flux can be considerable; for grassland and potatoes the reduction was 17 and 46% (Table 5) or 525 63 and 34 mm (Table 4).
About half of the yield increases was caused by internal recirculation as occurs in freedrainage conditions and the other half was caused by an increased upward capillary flow from groundwater. Improved modelling should consider upward flow of soil water which will result in improved estimates of crop yield and percolation.

530
We think that the quantification of recirculation on yield is a novelty, We also thank anonymous reviewers for their constructive and valuable comments on earlier versions of this paper. De Wit, A.J.W., van Diepen, C.A. (2008). Crop growth modelling and crop yield forecasting using satellite-derived meteorological inputs. International Journal of Applied Earth Observation and Geoinformation, 10(4): 414-425. De Wit, A.JW., Duveiller, G., Defourny, P. (2012). Estimating regional winter wheat yield with WOFOST through the assimilation of green area index retrieved from MODIS observations. Agricultural and Forest Meteorology, 164: 39-52.
De Wit, C. T. (1978). Simulation of assimilation , respiration and transpiration of crops.       Condition a is artificially created to explicitly demonstrate the role of recirculating percolation resulting in upward flow to the root zone. Condition b is a common free drainage situation which includes upward flow due to recirculating percolation water. Condition c is the natural situation in most of the Netherlands. This hydrological condition has a fluctuating groundwater level derived from a national study (Van Bakel et al., 2008).