The comments of Fu et al. (2026) on the paper of Rickard et al. (2025a) highlight a critical gap between the complexity of water–root–soil interactions in the field and our current capacity and ability to measure and model them. Given that New Phytologist is a plant journal, the focus of Rickard et al. (2025a) was the physiological perspective: how plants respond physiologically to spatiotemporal variations in soil water content under field conditions, while the hydrological perspective, including the method for calculating daily root water uptake, was only briefly explained (Rickard et al., 2025a). We take this opportunity to elaborate on this. In this response, in addition to addressing the specific points raised by Fu and co-authors, we discuss the complexity of soil–water–root interactions in the field, as well as their broader implications for what we can measure and model under field conditions. The premise of Fu and co-authors is that the Richards' equation and other methods used in their comments provide error-free solutions, which can then be used as a benchmark to verify our method. This premise is flawed in both theory and practice: the Richards' equation is only an approximation, and its errors for modeling water flow in field soils are unpredictable. Herein, we discuss the mechanisms underlying these errors and clarify why the method proposed by Rickard et al. (2025a) represents one of the most practical approaches for calculating daily root water uptake in the field, as noted by Burks Whalley et al., 2004, 2005; Rabbi et al., 2018; Wang et al., 2020), as well as their consequences for root water uptake in three dimensions, with 3D root architecture explicitly resolved (Zhang et al., 2020). We have decades of experience developing and applying numerical solutions of the Richards' equation and were aware of its limitations in modeling water flow under field conditions (Zhang Zhang et al., 2002, 2020; Huang et al., 2023). The method proposed by Rickard et al. (2025a) builds upon this body of work. Since Fu and co-authors used the Richards' equation as a benchmark to assess the method of Rickard et al. (2025a), we first discuss this equation and explain its problems in modeling water flow in the field. However, before proceeding further, we identify and address two errors in the comments made by Fu et al. regarding the method of Rickard et al. (2025a). The first error is that their comments misinterpret (via their statement ‘integration of the Richards' equation over········’) eqn (1) in Rickard et al. (2025a) as the Richards’ equation. In the literature, the Richards’ equation specifically refers to the equation derived by Richards (1931), in which Darcy's law, originally formulated for water flow in saturated soils, was extended to unsaturated conditions. Although the Richards’ equation conserves mass, it differs fundamentally from eqn (1) in Rickard et al. (2025a), which is a general mass-balance equation and does not require water flow to obey Darcy's law. This difference is critical when assessing the suitability of the Richards’ equation for modeling water flow under the conditions addressed by Rickard et al. (2025a). The second error relates to their claim that there is a dimensional mismatch in eqn (3) in Rickard et al. (2025a), without specifying the nature of the alleged mismatch. We have carefully re-examined the derivation of eqn (3) and can confirm that this equation is dimensionally consistent. In their comments, Fu and his co-authors use the MOIST model they developed based on the Richards' equation (Fu et al., 2025) to demonstrate what they interpret as errors in the results of Rickard et al. (2025a). The Richards' equation is an approximation rather than an exact description of the processes that govern water flow in soil; it was derived based on two key assumptions (Richards, 1931). The first one is that air pressure is spatially uniform across the soil profile, such that air flow and water flow can be decoupled and do not affect each other; this assumption may fail during rainfall events, during which infiltrating water isolates air into air bubbles or traps air into poorly connected pores, thereby influencing liquid water flow. This is a common phenomenon in water infiltration and can be easily tested using soil column experiments. The second one is that the Richards' equation requires soil water retention curve and soil hydraulic conductivity functions. These two parameters are typically measured in the laboratory under conditions in which water distribution in the pore space reaches an equilibrium state. Physically, this means that when wetting a dry soil, water cannot fill pores unless all smaller connected pores are already filled, while when drying a wetted soil, water in pores cannot drain unless all larger connected pores have already drained. Furthermore, the water retention curve obtained during drainage differs from that obtained during wetting, a phenomenon known as hysteresis (Basile et al., 2003). This equilibrium assumption breaks down during rainfall or irrigation events, during which infiltrating water flows preferentially through large and interaggregate pores, while bypassing smaller intra-aggregate pores, even when these pores remain air-filled. This nonequilibrium flow, also known as preferential flow and macropore flow in the literature, has been well-documented (Gerke Ewen, 1996; Beven this allows the derivation of average water retention curves for soils at different depths. However, due to technological limitations, current tensiometers cannot measure soil matric potential more negative than −300 kPa (Whalley et al., 2007). Second, soil physical properties are not static but highly dynamic, varying with time. Biologically, as plants grow, roots and root-associated processes, such as root exudation and microbial and faunal activity, reshape soil structure and alter the surface tension and viscosity of soil water over timescales of days (Crawford et al., 2012); these changes in turn modify the soil water retention curve and hydraulic conductivity (Read Whalley et al., 2004, 2005; Kroener et al., 2014; Schwartz et al., 2016; Choudhury et al., 2018; Rabbi et al., 2018; Helliwell et al., 2019). Physically, wetting and drying cycles caused by rainfall, irrigation and root water uptake result in soil shrinking and swelling, further modifying soil structure and the capacity of soils to retain and transmit water (Utomo Müllers et al., 2023). Recent discovery by Rickard et al. (2025a) and others (Müllers et al., 2023) that plants take up the topsoil water preferentially adds further complexity to this process. All of these observations indicate that root water uptake is not solely determined by root length, as assumed in Fu et al. (2025, 2026), but emerges from a complex root–soil interaction (Vadez, 2014), which cannot be described by the Feddes' model used by Fu et al. (2025). Mathematically, Fu et al. also overlooked that their MOIST model is formulated using the h-form of the Richards' equation, where the matric potential is used as the dependent variable. It has been well-established that numerical solutions of the Richards' equation in this form are not mass-conservative. Depending on spatial and temporal discretization in numerical solution and soil hydraulic properties, it can give rise to substantial errors (Celia et al., 1990). Taken together, these numerical errors, methodological oversights, inherent uncertainties and numerous processes that the Richards' equation cannot describe seriously undermine the suitability of the model of Fu et al. (2026) for mechanistically describing water flow in field soils. Fu et al. criticize our method for using night-time ∂Q/∂z to estimate daytime ∂Q/∂z, labelling it a ‘major assumption’. Mathematically, this is an approximation rather than an assumption; it is an approach used in all numerical modeling. In fact, when formulating their numerical solution, Fu et al. (2025, 2026) also assume that soil matric potential varies linearly within each time step, just as in our method, thereby introducing truncation errors. The only difference is that the time-step size in the method of Rickard et al. (2025a) is c. 12 h. Because such truncation errors depend on time-step size, Fu et al. might argue that their numerical solution can use smaller time steps (e.g. 10−8 s) to reduce such errors. While this is mathematically correct, field data, such as rainfall and other meteorological data (required for calculating evapotranspiration), are typically recorded daily or collected on event basis (as in the archived Rothamsted meteorological data used by Fu et al. in their comments). Disaggregating these recorded values into shorter intervals (e.g. 10−8 s) is purely a mathematical manipulation rather than a physically meaningful representation of rainfalls in the field. The errors induced by such approaches are uncertain and unpredictable. Consider a recorded daily rainfall event as an example. If the rainfall occurs during the preceding night of a sunny day, the rainfall enhances plant transpiration in the following day due to increased soil moisture; by contrast, if the same amount of rainfall occurs during the day rather than at night, the rainfall suppresses plant transpiration due to reduced sunlight and decreased vapor pressure deficit. That is, although the recorded daily rainfall in the two scenarios is identical, its consequences for soil water dynamics differ markedly. Therefore, when comparing with measured data, using a time step of 10−8 s instead of 12 h does not necessarily improve modeling accuracy. In other words, mathematical refinement of a numerical solution does not necessarily translate into improved modeling accuracy when limitations of available field data are considered. While rain gauges in modern weather stations can record a rainfall event at fine temporal resolutions, these measurements cannot be directly extrapolated to other locations due to spatial variability of rainfall intensity (will be discussed later). Fu et al. suggest combining the Richards' equation with Bayesian inference to estimate root water uptake, apparently unaware that we have done this (Huang et al., 2023). Calculating root water uptake using the Richards' equation and measured soil water data is an inverse problem, which is ill-posed, meaning that a small error in soil water measurements could lead to large errors in the calculated root water uptake. Due to the unmeasurable spatiotemporal variability of soil physical properties and the inherent inaccuracies of soil moisture measurement devices, inverse modeling is extremely challenging. In fact, Guderle the outcome deteriorates further when soil physical properties are not accurately known. While all models require high-quality data, inverse models based on the Richards' equation tend to amplify measurement errors in the estimated root water uptake. Fu et al. attempted to estimate the upper and lower bounds of root water uptake in the 0- to 50-cm soil layer, but their approach contains flaws. We explicitly stated (Rickard et al., 2025a) that our method cannot distinguish between evaporation and root water uptake. Accordingly, we carefully selected time periods during which evaporation was negligible and provided photographic evidence in the Supporting Information (Rickard et al., 2025a) to illustratively support this. In estimating the upper bound using rainfall and soil water storage changes, Fu et al. failed to account for canopy interception and overlooked that mounds formed by root expansion and soil fauna at different locations and times can induce surface run-off at spatial scales relevant to our experimental measurements. They also implicitly neglected (due to the assumption of zero matric-potential gradient at the bottom end) upward water flow from the subsoil (> 50 cm), which could become significant when the topsoil is dry. In their lower bound calculation, they included an evaporation term that did not exist and did not clarify how water percolation out of the 0- to 50-cm soil layer was calculated when the topsoil was wet. Since the leaf area index during the selected period is > 3, canopy interception is substantial. Indeed, because of canopy interception, not all rainfall events during this period resulted in measurable changes in soil water content. Furthermore, soil water content was measured at three depths (15, 30 and 45 cm), and Fu et al. did not give the inter- and extrapolation methods they used to estimate soil water storage changes in the 0- to 50-cm profile, which could significantly influence results. Beyond these issues, there is an additional problem in their use of the archived rainfall data in Rothamsted. The weather station at Rothamsted is located c. 200 m east of the experimental site. Although we do not have direct evidence of spatial variability of these rainfalls at Rothamsted, data we observed from other locations show significant variability. As an illustrative example, Fig. 2 shows the spatial variation of a rainfall event measured by twelve 2 m × 2 m weighing lysimeters at the Experimental Station of the Chinese of All these to the by Fu et al. not all root water uptake values in the 0- to 50-cm soil layer calculated using our method the upper bounds estimated by Fu et al., further that these from variability and other uncertainties discussed previously, rather than from our method. The method proposed by Rickard et al. (2025a) is based on and is It requires only soil moisture data and for the influence of numerous physical and processes, as discussed previously, that cannot be measured in situ and cannot be described by the Richards' equation. It is more than the Richards' equation for calculating daily root water uptake in the field such an approach is practical approach to our method is through weighing which provide direct daily measurements of We are currently such measurements at the Experimental Station of the of Chinese of We at the of Chinese of for lysimeter This work is by the & Station of in and by the and in the The New Phytologist with to in and in
Rickard et al. (Fri,) studied this question.