Observed Soil Water Content Research Articles

Deriving catchment-scale water and energy balance parameters using data assimilation based on extended Kalman filtering

Important catchment-scale water and energy balance parameters are derived for a small catchment in southeastern Australia by assimilation in a catchment-scale soil-vegetation-atmosphere transfer (SVAT) model of subcatchment-scale soil water content observations and land surface temperature measurements. In order to incorporate the subcatchment-scale soil moisture variability and its time evolution in a data assimilation scheme, an extended Kalman filter (EKF) method is used in combination with a cost function minimization approach to derive effective parameters for the catchment as a whole. These parameters are the minimum surface resistance to evaporation and the soil hydrodynamic parameters. This method provides a balanced assessment of all the uncertainties regarding the description of the catchment hydrological behaviour. Moreover, these uncertainties are propagated forward in time in a single framework, which combines soil moisture correction with effective parameter estimation. Two issues are addressed in this paper: (a) the applicability of the method for scaling purposes (effective parameterization) and (b) the relationship between effective parameters obtained by this method and parameters obtained by a classical minimization routine that ignores soil moisture correction and observation uncertainty. Effective parameters are found to be consistent with the available local parameter measurements. Derived parameter sets with and without EKF have been found to be very similar and this has been explained in a number of ways.

Modelling the Soil Moisture Deficits Developed under Grass and Deciduous Woodland: the Implications for Water Resources

The area of deciduous woodland in the UK is increasing as a result of Government policy, and studies suggest that this might result in reduced streamflow and groundwater recharge. Measurements of soil water content were carried out (a) at a site on a sandy loam soil, (b) under land covers of grass, and (c) in deciduous woodland. The data show that both the field capacity and wilting point of the soils were much less under the woodland than those under grass. In 1997, soil moisture deficits were developed down to a depth of 1 m under grass and in excess of 3 m under the deciduous woodland. A simple daily water-balance model was able to simulate the observed soil water contents and emphasised the importance of interception losses from the deciduous woodland canopy. For a 25-year period, the model predicted that the mean annual runoff and soil drainage for the woodland was less than half those for grass.

Evaluation of the root zone water quality model for predicting water and NO 3–N movement in an Iowa soil

Evaluation of computer models with field data is required before they can be effectively used for predicting agricultural management systems. A study was conducted to evaluate tillage effects on the movement of water and nitrate–nitrogen (NO 3–N) in the root zone under continuous corn ( Zea mays L.) production. Four tillage treatments considered were: chisel plow (CP), moldboard plow (MP), no-tillage (NT), and ridge-tillage (RT). The root zone water quality model (RZWQM: V.3.25) was used to conduct these simulations. Three years (1990–1992) of field observed data on soil water contents and NO 3–N concentrations in the soil profile were used to evaluate the performance of the model. The RZWQM usually predicted higher soil water contents compared with the observed soil water contents. The model predicted higher NO 3–N concentrations in the soil profile for MP and NT treatments in comparison with CP and RT treatments, but the magnitude of simulated NO 3–N peak concentrations in the soil profile were substantially different from those of the observed peaks. The average NO 3–N concentrations for the entire soil profile predicted by the model were close to the observed concentrations except for ridge tillage (percent difference for CP=+5.1%, MP=+12.8%, NT=+18.4%, RT=−44.8%). Discrepancies between the simulated and observed water contents and NO 3–N concentrations in the soil profile indicated a need for the calibration of plant growth component of the model further for different soil and climatic conditions to improve the N-uptake predictions of the RZWQM.

Predicted drainage for a sandy loam soil: sensitivity to hydraulic property description

Prediction of unsaturated flow phenomena at field-scale requires a set of hydraulic functions that capture local-scale variability which is present in all natural soils. Several sets of hydraulic functions measured on different core sizes collected across the field were used to predict drainage from a saturated soil profile. Simulations were carried out with a one-dimensional numerical model, based on the Richards equation. Four methods were considered: (1) Monte-Carlo simulation using 500 unimodal retention and conductivity functions representing θ(ψ) and K s data measured on 0.05 m diameter 0.051 m long core samples; (2) multiple simulations with a set of 60 multimodal retention and conductivity functions which better represented the measured θ(ψ) data of method 1; (3) one single simulation with a set of hydraulic functions obtained from a gravity-drainage experiment on 15 1 m long 0.3 m diameter soil columns collected from the same field; (4) one single simulation with a set of hydraulic functions obtained by scaling the retention and conductivity data from method 3. A perfect match of the final mean outflow was obtained when scaled retention data was used in combination with scaled K s values (method 4). All other cases underestimated the total outflow to a varying degree: 30% for method 1, 29% for method 2, and 21% for method 3. The results further revealed that none of the four methods was able to completely describe the mean observed drainage from the start until the equilibrium condition. This was further demonstrated by the disparities between the mean observed soil water content profile and the simulated values using hydraulic functions from method 4. Especially after the first day, differences were large, presumably because macropore flow could not be described using the Richards flow equation. Despite the introduction of multimodal retention and conductivity functions which better described the retention behaviour of small soil cores (method 2) in comparison with unimodal retention functions (method 1), mean predicted outflow for both methods was nearly identical.

Applying stochastic simulation techniques to a deterministic vadose zone solute transport model

Current models for simulation of water flow and solute transport through the vadose zone are limited in their applicability to field-scale problems. On the one hand, most of the models lack the capability to account for naturally existing soil spatial heterogeneity and complex (bio)chemical processes that many solutes undergo, and on the other, use of certain types of the existing models is often not economically feasible. In the present study, stochastic simulation techniques were applied to GLEAMS, a deterministic vadose zone solute transport model, to simulate soil water changes and nitrate (NO 3 −) concentrations at various depths in the top 1.2 m heterogeneous soil profiles of three research plots in southern Ohio. Simulations were conducted at three levels of complication: (1) single-value output was obtained for each plot by running GLEAMS with plot-based mean values of soil parameters; (2) a prediction band was formed from the output probability density function (pdf) with selected soil properties described by a multi-variate normal (MVN) random vector for each plot; and (3) with each plot decomposed into sub-environments, a prediction band was formed from all output pdfs obtained using MVNs for every sub-environment. Predicted and observed soil water contents and nitrate concentrations were compared to evaluate model performance. Overall, the predicted long-term means from all simulations fit the long-term means of field observations well. However, stochastic simulation at the sub-environment level showed two distinct merits over the other two approaches: (1) it better predicted variations in soil water content and nitrate concentration at a plot scale; and (2) it had the capability of locating high-risk areas at a sub-environment scale.

Observed and Simulated Solute Transport Under Varying Water Regimes: I. Bromide and Pentafluorobenzoic Acid

AbstractComparison of observed and predicted solute transport under field conditions is a necessary component for validating various modeling approaches. The objectives of this study were to compare transport of Br− and pentafluorobenzoic acid (PFBA) under cropped and fallow conditions for a range of soil water contents and evapotranspirative demands, and to evaluate the predictive capability of a computer simulation model (LEACHM) using independent estimates of model input parameters. A field study was conducted (1991) using in situ columns (0.2 m diam., 1.22 m depth) on a Brocko silt loam (mixed, mesic Borollic Calciorthid) in Gallatin Co., MT. The transport of two tracers (Br− and PFBA) was measured at three depths (0.36, 0.66, and 0.96 m) for three water regimes under cropped (barley, Hordeum vulgare L.) and fallow conditions for 60 d following chemical application. Observed breakthrough curves (BTCs) demonstrated increased travel times with increasing depth and decreasing water application. Solute transport in cropped columns was significantly slower than fallow as a result of higher evapotranspiration. Comparison of Br− and PFBA BTCs under cropped conditions suggested significantly greater plant uptake of Br− compared to PFBA. Predicted volumetric water contents, evaporation (under fallow conditions), and solute BTCs using LEACHM agreed favorably with observed results. Under cropped conditions, agreement between predicted and observed soil water contents was less favorable, most likely a result of inaccurate assumptions concerning depth‐dependent root water uptake. LEACHM‐predicted mean travel times (d) agreed within 22% of observed PFBA values for fallow conditions and within 9% of observed PFBA values for cropped conditions. Results indicated that adequate predictions of the transport of PFBA were obtained using LEACHM when coupled with independent estimates of model input parameters.

A simulation of the surface energy budget and soil water content over the Hydrologic Atmospheric Pilot Experiments‐Modélisation du Bilan Hydrique forest site

The surface energy budget and soil water content variation are simulated for a two month period during the 1986 Hydrologic Atmospheric Pilot Experiment‐Modélisation du Bilan Hydrique (HAPEX‐MOBILHY) (J.C. André et al., 1988) using the Oregon State University Coupled Atmosphere Plant Snow (CAPS) model. When initialized and driven by the observed soil water content and atmospheric forcing, the CAPS model closely simulated the observed surface energy budget and the variation of soil water content. Lateral transports of soil water may cause significant impacts on the simulated soil water content during the later part of the simulation. Vegetation appears to significantly affect the surface energy budget and temporal variation of soil water content. We found two major mechanisms by which vegetation controls the surface energy budget and soil water content: (1) dependencies of transpiration on ambient conditions, and (2) direct hydrological coupling between the atmosphere and deep soil water through the root structure. Compared to the simulation without vegetation, net effects of vegetation are (1) sustained evaporation over the entire simulation period, (2) more gradual variation of near surface soil water content, and (3) continued reduction of deep soil water content through transpiration. Soil heat flux becomes an important component of the surface energy budget over a bare soil surface. The simulated surface energy budget is strongly dependent on the initial soil water content, however, dependence of the simulated surface energy budget on the initial soil temperature was relatively minor.

Teste de um modelo de monitoramento de água no solo para uma cultura de sorgo submetida a diferentes tratamentos de irrigação

Um modelo de balanço hídrico diário utilizando informações de estação meteorológica automática, fenologia e informações edáficas foi ajustado e testado para uma cultura de sorgo usando experimentos de campo com diferentes tratamentos de irrigação durante o verão de 1990 e 1991, em Mead, Estado de Nebraska-EUA. Estimativas do total de água no solo a partir do balanço hídrico compararam-se bem com as leituras de sonda de nêutrons tomadas nos diferentes tratamentos. O desempenho do modelo, por camadas de solo, indicou pequena subestimativa da umidade nas camadas superiores, pequena superestimativa nas inferiores e boa estimativa nas intermediárias. A eliminação desses erros resultaria em melhor desempenho do modelo nas diferentes camadas. Boas estimativas do total de água no solo podem ser obtidas através deste balanço hídrico edafoclimático modificado com base em informações fenológicas, edáficas e de dados obtidos de estações meteorológicas automáticas.

Modeling the physical environment of tree seedlings on forest clearcuts

A soil and atmospheric boundary layer model has been developed to predict soil temperature, soil moisture, air temperature and humidity on forest clearcuts. The model comprises a numerical solution to the one-dimensional transfer equations for energy and water in the air and soil near a surface. Heat and water transport in the soil are coupled using a linked source-term approach. Model predictions were compared with field observations from three forest clearcuts in northern Ontario. Soil temperatures on all sites were predicted with less than 1°C bias. It was necessary to increase the value for soil hydraulic conductivity in the model to obtain good agreement between predicted and observed soil water contents. During lapse conditions on two sand flat sites, air temperature at 20 cm was predicted with less than 0.1°C bias and water vapor pressure at 20 cm with less than 10 Pa bias. As a consequence, the 20-cm saturation deficit was predicted with less than 10 Pa bias. On a swamp site, however, bias in predicted air temperature was much larger, possibly because of heterogeneity of the surface. In most instances, uncertainty in observed or predicted values was of similar magnitude to the differences between predicted and observed values. The model is suitable for examining the effects of soil and surface properties and of boundary conditions on the physical environment of flat, homogeneous, unvegetated forest clearcuts.

Soil water content at R-5. Part 2. Impact of antecedent conditions on rainfall-runoff simulations

In this paper, the second part of a two-part series, the performance of a quasi-physically based rainfall-runoff model (QPBRRM) is evaluated for a set of 39 events from a small rangeland catchment (R-5) using a suite of alternative antecedent soil water content estimates. The data set used to glean antecedent conditions for the R-5 events, described in the companion paper contains 22 848 observations of soil water content made over a 4 year period. The performance of the model is improved by using estimates of antecedent soil water content that include consideration of spatial variability.

Measurement and simulation of evaporation from a red earth. II. Simulation using different evaporation functions

The performance of several alternative evaporation functions for simulating water loss from a bare red earth was assessed by including them in a computer model which was used to simulate evaporation both from red earth monoliths in a glasshouse, and from a study site in the field. The coefficients in the different evaporation functions were also optimized to minimize the root mean square discrepancy (RMSD) between simulated and observed soil water contents. RMSD values for the alternative evaporation functions before optimization of coefficients ranged from 3.2 to 7.0 mm for the glasshouse data and from 4.0 to 6.6 mm for the field data. Optimization reduced these values 3.0 to 6.4 mm (glasshouse) and 3.9 to 6.1 mm (field). The sensitivity of the model to errors in hydraulic conductivity estimates was assessed. Overestimating hydraulic conductivity by 2 and 10 times increased predicted cumulative evaporation by 8 and 28% respectively. Underestimating conductivity by the same factors produced similar reductions in predicted cumulative evaporation. The model was used to test the effect of basing the simulation of field evaporation on different thicknesses of surface compartment, for two alternative evaporation functions. Optimum thicknesses of surface compartment were 20 and 30 cm, and increasing these thicknesses to 60 cm resulted in only c. 20% increase in RMSD. This effect was considerably less than the increase caused by using inferior alternate types of evaporation function.