In many regions of the world, plant growth and productivity are limited by water deficits. As a result of more frequent and intense droughts, the area of land characterized as very dry has more than doubled since the 1970s. Consequently, understanding root water uptake under water-stressed conditions is gaining importance. The performance of a recently developed polymer tensiometer (POT) designed to measure matric potentials down to −1.6 MPa was evaluated and compared with volumetric moisture content measurements in dry soil. Three irrigation intensities created severe, intermediate, and no water stress conditions in lysimeters with growing maize (Zea mays L.) plants. By monitoring matric potentials using POTs, levels of local water stress in our experiments were better defined. When the defined stress levels were reached, volumetric moisture measurements for this particular loam soil were below 0.1, thus less informative compared with matric potential measurements. The observed matric potential profiles indicate significant root water uptake between 0.3- and 0.5-m depth in the later growth stages under water-stressed conditions. The temporal pattern of matric potential profiles reflected changing root water uptake behavior under dry conditions. As the total soil water potential is a direct indication of the amount of energy required by plants to take up water, POTs may contribute to quantifying root water uptake in dry soils.

Direct soil water potential measurements by polymer tensiometers help in defining levels of local water stress in root water uptake experiments, especially in dry soil where moisture content measurements become less informative. Observed temporal matric potential patterns showed changing root water uptake behavior under water-stressed conditions.

Water deficits limit plant growth and productivity in many regions of the world. As droughts have become more frequent and intense, the area of land characterized as very dry has more than doubled since the 1970s (Dai et al., 2004; Huntington, 2006). Although 70 to 85% of the world's consumable water is currently allotted to agriculture, increasing urban demands for potable water due to population growth will continue to compete with agricultural water use from now on (Somerville and Briscoe, 2001; Foley et al., 2005). The need to understand plant responses to water deficits has never been more acute. The driving force for root water uptake is the water potential gradient between root and soil. Unsaturated water flow in the root zone often occurs at soil water matric potentials that are below the range of water-filled tensiometers (to approximately −0.09 MPa; Young and Sisson, 2002). As a consequence, little is known about the distribution of root water uptake throughout the root zone under dry conditions. Hopmans and Bristow (2002) and Feddes and Raats (2004) noted that the lack of knowledge about root water uptake under stressed conditions has unavoidably led to the rather schematic representation of the root zone in unsaturated models.

Long-term water stress alters the physiological functioning of the entire plant (Kramer, 1983; Smith and Griffiths, 1993), but little is known about the short-term dynamics of root water uptake under water-stressed conditions. New techniques such as x-ray tomography, nuclear magnetic resonance, and two-dimensional light transmission imaging have improved our understanding of plant responses to dry conditions (Tollner et al., 1994; Van der Weerd et al., 2001; Garrigues et al., 2006), but these techniques are complicated and not suitable for field use. Fortunately, the recent development of a POT (Bakker et al., 2007) that measures soil water matric potentials to −1.6 MPa creates new possibilities for studying the dynamics of soil water matric potentials in the vicinity of roots, in the laboratory as well as the field.

The total soil water potential, consisting of the matric, osmotic, gravitational, and other potentials, is a direct indication of the amount of energy required by plants to take up water. A plausible hypothesis is that a plant distributes the water uptake throughout its root network in a way that minimizes the expenditure of energy at any given moment (Dirksen et al., 1994; Adiku et al., 2000). This would be consistent with the observations of Bohm et al. (1977) that root water uptake and root density correlate poorly. Direct observation of low soil water matric potentials within a plant's root zone, however, may provide knowledge of the plant's ability to satisfy its water needs (and of its strategy to maximize this ability).

The main objective of this study was to evaluate the performance of POTs in the root zone during an entire cropping cycle, including the POT's ability to register matric potentials across the full range encountered in a cropped soil. A second objective was to investigate the added value of soil water matric potential measurements for root water uptake studies under water-stressed conditions. The dynamics of soil water matric potential and soil moisture were investigated under varying levels of water stress in three cropped lysimeters.

Materials and Methods

Plant water stress will increase with decreasing matric potentials until matric potentials reach a level where root water uptake is no longer possible. For agricultural crops, this level is around −1.6 MPa (Koorevaar et al., 1983); however, the matric potential range within which different crops begin to be affected by water stress is less well known.

Taylor and Ashcroft (1972, p. 434–435) provided matric potential data at which “water should be applied for maximum yields of various crops grown in deep, well-drained soil that is fertilized and otherwise managed for maximum production” including data for maize. Kroes and van Dam (2003) used these matric potential values in combination with the water stress reduction function of Feddes et al. (1978). Based on the Feddes reduction function and the data of Taylor and Ashcroft (1972), we defined three treatments: no stress (NS, minimum pm = −0.15 MPa), intermediate stress (IS, minimum pm = −0.45 MPa) and severe stress (SS, minimum pm = −0.80 MPa), with minimum pm being the most negative matric potential (measured by any of the POTs) that was allowed during the specific treatments.

We constructed three lysimeters of 0.7 by 0.5 by 1.7 m high (Fig. 1 ). Each lysimeter consisted of three vertical segments of 0.5-m depth for easier backfilling. Below the lowest compartment, a drainage compartment of 0.2-m depth was filled with gravel to support the weight of the soil. Each lysimeter was placed on a stainless steel frame to allow weighing on a movable scale (1500-kg range, 0.2-kg accuracy). We filled the lysimeters with a prewetted loamy soil (10.9% clay, 58.2% silt, 31.9% sand) in 0.05-m layers, packing the layer by tamping with constant force, and raking the upper 0.02 m before applying a new layer. We installed three instrumented layers at 0.2-, 0.6-, and 1.1-m depths during filling. Each contained three POTs (Bakker et al., 2007; measurement range 0 to more than −1.6 MPa, accuracy 2.38 × 10−3 MPa) and three time domain reflectometry (TDR) probes (three wire, 0.1 m length, 0.0175-m wire spacing, observed volumetric moisture content resolution of 0.001 [−]). Instead of porous ceramic plates (Bakker et al., 2007), we outfitted the POTs with solid ceramic cones of the same material. The POTs were calibrated for long-term pressure decay and temperature influence according to Bakker et al. (2007). Preliminary testing showed that the distance between each POT and TDR probe was sufficient to prevent measurement interference (see also Baker and Lascano, 1989; Zegelin et al., 1989). Because root profiles of maize may be greatly altered by temporary drought stress (Box et al., 1989), we monitored root growth during the experiment by installing horizontal acrylic rhizotubes every 0.1-m depth, except at 0.5 and 1.0 m due to the construction of the boxes. The acrylic material itself does not have an effect on root growth (Brown and Upchurch, 1987; Johnson et al., 2001). Root images were captured using a cold-light boroscope with a diameter of 6.35 mm (Heine Optotechnik GmbH, Herrsching, Germany) and a digital camera (Nikon Coolpix 4500). The final images were 320 by 240 pixels and captured a curved rhizotube area of 1 cm2. During the experiment, root presence was detected by visual inspection of the images.

The lysimeters were insulated by a layer of 0.1-m polystyrene (on the sides of the lysimeters) to reduce temperature influence on root growth (McMichael and Burke, 1996). The lysimeters were uniformly irrigated by emptying syringes with equal amounts of water in each of the cells of a metal grid firmly placed on the soil surface. Because nutrient analysis indicated that the soil contained enough nutrients for plant growth during the experiment, unchlorinated tap water was used for irrigation instead of nutrient solution. Each lysimeter had an artificial growing light and reflective material to enhance plant growth in the lysimeter. The reflective material restricted light interference from the other lysimeters, ensuring that each lysimeter received an equal amount of light. Before sowing, the soil in the lysimeters was left to consolidate for a month, and irrigated to maintain realistic moisture profiles.

On 5 Sept. 2005, maize was sown in every lysimeter at the horizontal coordinates (x, y) = (0.25, 0.117 m), (0.25, 0.35 m), and (0.25, 0.467 m) [with (0, 0) denoting the back left corner, x running from back to front, y from left to right) to ensure that equal volumes of soil were available for each maize plant. The plants were optimally irrigated (Fig. 2 ) during the initial growth stage of approximately 20 d (Allen et al., 1998). After 25 d, we terminated irrigation on two lysimeters until the prescribed water stress level was attained at one of the shallow POTs. Minimum pm was reached on 24 Oct. 2005 for the IS, and on 2 Dec. 2005 for the SS treatments. The stress level was subsequently maintained by using small, frequent irrigations. The remaining lysimeter was irrigated throughout the experiment to maintain the minimum pm as the no-stress treatment. Temperature and relative humidity (RH) were recorded with a thermohygrometer (Oregon Scientific, Portland, OR; accuracy 0.1°C and 1% RH) above one lysimeter for the first 25 d during which the irrigation was the same for all treatments, and thereafter above every lysimeter on alternating days to record possible changing conditions between the lysimeters. The thermohygrometer was placed on the polystyrene insulation, close to the soil surface but not hindering evaporation. The RH data were converted to vapor pressure deficit (VPD) (Allen et al., 1998). The experiment lasted 4 mo until all plants completed their growth cycle. At this point, undisturbed soil samples were taken to determine soil hydraulic properties (98-cm3 sample rings) using the soil core method (Blake and Hartge, 1986). Samples taken close to the TDR probes were used for gravimetric water content determination at 0.2, 0.6, 1.2, 3.2, and 10 kPa (averaged across the 0.05-m-high sample) on a suction table (Romano et al., 2002). Disturbed samples were used in a pressure plate extractor setup (Dane and Hopmans, 2002) to determine the matric potential–soil moisture relation at 0.1 and 1.6 MPa. Root length density (RLD) samples were obtained using an auger with plunger and a sample volume of 385 cm3 (Oliveira et al., 2000). The RLD samples were hand washed according to Oliveira et al. (2000) to remove the soil, stored in a 10% ethanol solution for a maximum of 3 wk, and scanned at 400 dpi and analyzed with the computer program WinRHIZO (Himmelbauer et al., 2004). We determined a cumulative RLD (m m−2 soil), which is the length of root present under a unit area of soil surface to a specified depth (Atkinson, 2000).

Results and Discussion

Immediately after filling, the mass of the lysimeters differed by <3.2 kg (see Table 1 ), indicating a uniform packing process. On the other hand, the soil bulk density profile determined after the experiment showed some variation within and between lysimeter soil profiles (Fig. 3 ). Possible reasons for this variability are unavoidable heterogeneity in packing or initial water content, shrinking or swelling under the different moisture regimes, and the nonuniform occurrence of roots in the soil profile. The CV is 0.03911 (n = 14) for the wet, 0.04203 (n = 15) for the intermediate, and 0.05525 (n = 15) for the dry treatments. According to Hillel (1998), bulk density distributions with a CV <0.15 can be considered highly uniform, and we consider our observed density variations as minor; however, even small variability in density may affect moisture distribution.

Lysimeter soil surface temperatures varied between 18.5 and 32.8°C and the VPD between 0.89 and 3.17 kPa. Table 1 shows the total irrigation and the evapotranspiration decrease with increasing stress levels (see also Fig. 2). Total drainage differed only slightly among the lysimeters (Table 1), but after Day 25 drainage ceased for the remainder of the experiment in all treatments.

Figure 4 shows distributions of matric potentials (left) and soil moisture (right) down the profile for all lysimeters at three dates. The contour plots were constructed from the nine POT and nine TDR observations in each lysimeter (see Fig. 1 and 4d). In the lysimeters, the TDRs were not located in one vertical plane, but in an L-shaped layout (top view, Fig. 1). For clarity we “unfolded” the L shape and plotted it in two dimensions, with the bend indicated by a dashed line at x = 0.25 m. The z line (vertical coordinate; zero at the soil surface, positive upward) at x = 0.2 m in the figure is located at (x, y) = (0.35, 0.25 m) in reality. The central maize plant in each lysimeter was sown at (x, y, z) = (0.25, 0.35, −0.05 m).

On 8 Sept. 2005 (Fig. 4a), all lysimeters were still receiving equal amounts of irrigation (see Fig. 2). The matric potential profiles were therefore still in the wet range and rather featureless. The profiles were wetter in the top of the lysimeter because of irrigation, and drier at the bottom as a result of the low moisture content at packing and the seepage face boundary condition. The minor differences in moisture content among the lysimeters can be explained by small differences in density, which affects the moisture distribution but not the matric potential. No roots were observed in the rhizotubes to this date. We therefore consider the observed moisture and potential profiles to be strictly abiotically controlled by soil hydraulic properties and the initial and boundary conditions.

By 23 Oct. 2005 (Fig. 4b), the IS and SS lysimeters had not been irrigated for 4 wk. The matric potential profile of the NS treatment remained in the wet range. In contrast, matric potentials in the IS and SS treatments were between −0.3 and −0.5 MPa in the upper profile, with similar matric potentials below 0.7-m depth. The volumetric moisture content in the NS treatment was still >0.1 m3 m−3 in most of the profile as a result of continued irrigation. The soil profiles in the other two treatments were completely (IS) or largely (SS) below 0.1 m3 m−3 volumetric moisture content. Rhizotube observations at this time showed that roots had grown to −1.2-m depth in the NS and to −0.8 m in the IS and SS treatment, indicating restricted root growth in stressed plants.

On 8 Dec. 2005 (Fig. 4c), the matric potential in the NS treatment remained essentially unchanged, but the other treatments indicated continued drying throughout the entire lysimeter. At this date, the SS lysimeter had a significantly lower matric potential than the IS lysimeter at all depths. Consistent with the matric potential data, the moisture contents in the SS lysimeter were somewhat lower than in the IS lysimeter, whose moisture distribution had hardly changed since 23 October because it was maintained at its target matric potential. All lysimeters had received irrigation a few days before, which resulted in slightly higher matric potentials in the upper part of the IS profile compared with 23 October.

With both the IS and SS lysimeters firmly in the dry range, the moisture content differed much less than the matric potential (compare moisture contents and matric potentials in the IS and SS treatments in Fig. 4c). The limited change in moisture can be explained by the shape of the water retention curve (Fig. 5 ). The observed matric potential and moisture content data from Fig. 4 were paired, coupling each POT to the nearest TDR probe, resulting in 27 water retention data points. These in situ water retention data were plotted together with the gravimetrically determined data from undisturbed samples and pressure plate extractor measurements. Almost all of the in situ water retention data have moisture contents <0.2. Moreover, the two stress treatments have their target potentials in the steep end of the retention curve, where the moisture content is <0.1, and thus the moisture content data provide little information due to the limited applicability of TDR in this dry range. Apart from representing the effect of different soil moisture regimes, the combined POT and TDR measurements also reflect the hysteretic nature of the soil water characteristic.

The gravimetric data in Fig. 5 clearly show the effect of small structure differences (a result of packing, swelling and shrinking properties of the soil, and root presence) on the moisture content. In wet soils (p > −10 kPa), where structure affects the moisture content, the CV of the moisture content ranged from 0.11 to 0.22 (n = 26). In contrast, in the dry end (p < −0.1 MPa), the CV ranged from 0.050 to 0.045 (n = 10), reflecting the dominating effect of soil texture under dry conditions.

The patterns depicted in Fig. 4 suggest that the horizontal distance to the stem has only a negligible effect on the effort required to extract water from the soil. Alternatively, one could argue that the bulk of the water was lost through the soil surface (evaporation); however, the evidence of drying at depth supports a significant role of the roots in removing water from the soil. Overall, the contour plots suggest a fairly uniform drying process laterally, with the nonuniformity mainly stemming from the spatial variation in initial wetting and soil properties. More importantly, Fig. 4 clearly shows that in the water-stressed cases after 23 Oct. 2005 (IS and SS treatment), the matric potential gradient between 0.4- and 0.6-m depths is larger than the gradient between 0.2- and 0.4-m depths, and that the water potential is larger (less negative) at depth. The upward Darcian flux density is therefore larger between 0.4- and 0.6-m depths than at 0.2- and 0.4-m depths, thus indicating that water must have been removed between those depths. Clearly, the only available mechanism is root water uptake. As the plants matured and the soil dried, water extraction progressed to greater depths, until, in the end, water at all depths could be potentially targeted by the roots.

Rhizotube observations showed root growth throughout the profile of all lysimeters. It proved impossible to quantify RLD from the rhizotube observations during the experiment. Collecting good quality images was difficult, and root growth along rhizotubes may not be comparable with roots growing in bulk soil. The matric potential and moisture profiles varied greatly at the end of the experiment (see also Table 1 for water storage differences at the end of the experiment). The cumulative RLDs that were determined from soil sampling did not show large variations (Fig. 6 ), with IS about 6 m m−2 less than the nearly similar values of NS (97.4 m m−2 at −1.4 m) and SS (103.6 m m−2 at −1.5 m). Limiting pm values were reached some time after ceasing irrigation, thereby suggesting a reason for the similarities in cumulative RLD irrespective of treatment. Stress levels, however, gradually increased during the experiment, and stress levels were reached well before the end of the growth cycle. Moreover, comparing treatments, the differences in cumulative RLD are more pronounced between the IS and the other treatments, although the stress levels of the IS and SS treatments were comparable during the first month after ceasing irrigation.

Coelho and Or (1999) suggested the measured RLD to be a measure of potential water uptake area, as was observed in kiwifruit [Actinidia deliciosa (A. Chev.) C. F. Liang & A. R. Ferguson] by Green and Clothier (1995). This seems to imply a limited effect of actual water stress on the root distribution and density. In this case, the root system may adapt to the soil water regime by developing fine roots and root hairs (which cannot be measured in RLD determinations) at preferred locations or depths to locally increase water uptake efficiency. Increased root water uptake efficiency of deep roots in combination with the capacity of maize to slowly grow roots in dry soil (Sharp and Davies, 1985) may have contributed to the observed similarities in cumulative RLD.

Conclusions and Recommendations

This study evaluated the performance of POTs in cropped soil and explored the potential for using POTs in lysimeters under varying levels of water stress. The POTs successfully monitored the soil water matric potentials of the two water-stressed treatments, thus providing a means to better define levels of local water stress. The water retention curve showed that volumetric moisture contents for this particular loam soil were <0.1 when water stress reached the defined stress levels. For these low moisture contents, TDR measurements may be of little use. Cumulative RLD data showed an inconclusive effect of the water stress levels; increased water uptake efficiency from deep roots in combination with slow but continuing root growth in drier soil may explain the observed similarities.

We conclude from the soil matric potential profiles given in Fig. 4 that, under water-stressed conditions after 23 Oct. 2005, significant root water uptake occurred between 0.3- and 0.5-m depths. The maize plants were able to extract water under very dry conditions, and continued to extract water from those dry regions. Such observations indicate that POTs have the potential to significantly improve experimental analysis of root water uptake, and thus they may help unravel plant root water uptake strategies under various levels of water stress.

This knowledge can improve the representation of the root zone in unsaturated models, and may ultimately be used to optimize irrigation regimes. Other potential applications lie in rainfed agriculture with generally much drier soils, and in rangelands or other unirrigated ecosystems.

This research was supported by the Technology Foundation STW, Applied Science Division of NWO and the technology program of the Ministry of Economic Affairs. We thank Peter van der Putten and Bart Timmermans of the Plant Sciences Department at Wageningen University for familiarizing us with RLD determination and letting us use their facilities. We are very grateful to Harm van Bentem for generously providing the soil we used in our experiment.

Freely available online through the author-supported open access option.