Interactive comment on “ Groundwater influence on soil moisture memory and land – atmosphere interactions in the Iberian Peninsula

I have added a combined pdf with the 3 responses to reviews as it was required in the "final response", but I encourage to use the responses posted here as author comments since they are easier to handle. We have produced the manuscript using online latex, therefore we do not have a tracked version in pdf comparing the original and final versions. The changed introduced are very clearly stated though, with quotes of the revised texts, in the author responses to reviews.


Summary Motivation
Climate change can only be assessed with Earth System Models. It is essential to understand how well the models represent the processes and feedbacks that take part in the climate system that couples the atmosphere, the ocean and the land surface. It is in this context, that this thesis attempts to contribute in the model representation of climate.
Groundwater is 30.1% of the total fresh water on Earth [107]. In the global hydrological cycle, water fluxes from soil storage to groundwater reservoir are estimated to be 46,000 km 3 yr −1 , and from the groundwater reservoir to the rivers and lakes 43,800 km 3 yr −1 [34]. These figures reveal the importance of the groundwater reservoir within the Earth fresh water. It is well known that groundwater interacts with the soil above, with surface water and with the ocean. But, 1) how does groundwater affect climate?, 2) is groundwater a player in land-atmosphere coupling?, 3) can groundwater interactions with the land-atmosphere system be represented in land surface models?, 4) in semiarid regions of the interior Iberian Peninsula, for instance, where land-atmosphere coupling plays an important role in precipitation recycling [93] and there are measured evidences of shallow water tables during dry years due to previous wet episodes [80], will a model representation of the groundwater interactions with the land-atmosphere system make a difference in climate simulations?
The working hypothesis is that YES answers Questions 2, 3 and 4. Testing the hypothesis and seeking an answer to Question 1 motivate this thesis work in the USC Non Linear Physics Group. A brief resume of the thesis is presented next:

Thesis structure and brief resume
Chapter 1 presents a general introduction in order to contextualise this research.
Firstly, a brief description of the natural water cycle and its main reservoirs and processes is given. The groundwater reservoir plays an important role in the natural water cycle; a) interacting with the unsaturated zone of the soil through gravitational drainage and capillary rise (groundwater recharge), b) transporting groundwater within the saturated zone, driven by topography (lateral groundwater flow), and c) interacting with the rivers and lakes, receiving infiltration and feeding the baseflow (groundwater-streams exchange). Secondly, an Iberian Peninsula climate variability description, including long-term precipitation analysis, introduces the area of study. The need to be cautious investigating hydrological processes in the peninsula is pointed out, dividing it into basins or hydrological zones and studying long-term periods, since the pluviometric regimes differ greatly within the peninsula and several-year long droughts occur all over it. Finally, an overview of recent efforts in assessing the groundwater link to soil moisture, land-atmosphere fluxes and surface streams concludes the chapter, focussing on the importance of considering groundwater dynamics in Iberian Peninsula modeling research.
In Chapter 2, the main tool used in this thesis to assess the role of groundwater dynamics in the Iberian Peninsula soil moisture and land-atmosphere fluxes is pre-  [63,64], and also represents the groundwater reservoir based on a dynamic groundwater scheme first presented by Miguez-Macho et al. (2007) [83], that has been continuously developed and reviewed since then as part of an ongoing collaborative work between the USC Non-Linear Physics Group and the Rutgers University Department of Environmental Sciences.
The first section of the chapter introduces LEAFHYDRO methodology for the land surface and the unsaturated soil layers (up to 4 m deep), consisting of the calculation of incoming and outgoing water and heat fluxes for each layer and surface represented, and the prognosis of temperature and moisture values through water xii mass and energy balances once the fluxes are known. A review of these calculations is presented, starting by the unsaturated soil formulation and following up through snowcover and temporary surface water, vegetation, bare ground, canopy air, precipitation and radiation. The second section of the chapter is a complete presentation of the LEAFHYDRO dynamic groundwater scheme, where the groundwater interactions and their model representation are detailed: two-way groundwater-soil fluxes, groundwater lateral flow and two-way groundwater-streams exchange. The chapter concludes with a description of the river flow scheme that closes the water cycle in the model.
In Chapter 3, the main experiment carried out in this thesis is described first. The experiment consists of two LEAFHYDRO 10-year period simulations with a 2.5 × 2.5 km grid; 1) the WT (Water -Table) run with the dynamic groundwater scheme switched on, that closes the Iberian Peninsula long-term water cycle accounting for the groundwater interactions with the soil and the land surface, and 2) the FD (Free-Drain) run with the dynamic groundwater scheme switched off, that uses the commonly adopted free-drain approach, where soil water is allowed to drain out of the soil column at a rate set by the bottom soil layer hydraulic conductivity.
Comparing the WT and FD runs, the influence of groundwater can be isolated. Then, the model initial topography, vegetation and soil parameters are shown. For the water table and soil moisture initialization, an initial Iberian Peninsula Equilibrium Water Table Depth estimation by Gestal-Souto et al. (2010) [47] is used, based on a two-dimensional groundwater model [42] that finds a balance between the atmospheric influence in the form of groundwater recharge and the topographic influence in the form of gravity-driven lateral convergence. The process to obtain a more realistic initial water table condition using the yearly groundwater recharge from a LEAFHYDRO 10-year test run is described, as is the calculation of the initial soil moisture conditions using Richards' Equation for soil water movement and the water table initial position. The initial river parameters origin and processing, as well as the use of surface atmospheric forcings from ERA-Interim reanalysis and Iberian Peninsula precipitation analysis (0.2 • spatial resolution), are detailed.
In Chapter 4, water table depth (td) and streamflow observational data are used to validate the LEAFHYDRO simulations and hence support the results in following chapters. The chapter is introduced with a discussion about the model parameterization of local lateral groundwater flow, which resolves subgrid internal drainage and results in a baseflow consistent with topography and climate, and not solely dependant on vertical fluxes or the water table position. A realistic water table spatial distribution and time evolution is essential to assess the groundwater-soil link, and model performance validation with observational td data at 623 stations shows; 1) the model captures shallow water table positions (td ≤ 8 m; greater connection to the land surface), as 66.0% of the observational shallow points are also shallow in the model; 2) the model captures the mean water table positions, as 33.0% xiii of the shallow water table stations present less than 2 m difference between simulated and observed td means (14.0% when all stations are considered); and 3) the seasonal cycle and long-term trends are realistically represented in the model, as 32.3% of the observed time series present a correlation coefficient greater than 0.5 with the simulated time series.
The representation of streamflow, as the last player in the water cycle discharging ultimately to the ocean, closes the water budget in the model. From validation with observational streamflow data and comparison between the FD and WT runs, the streamflow seasonal cycle is found to be differently represented by both runs; winter higher and closer to observations streamflow in the FD run, better representation of summer baseflow in the WT run, and faster recovery of streamflow when the autumn precipitation begins (September-October) following observations in the WT run. Correlation coefficients between observed and simulated streamflow time series indicate the better simulation of the seasonal cycle and interannual variability by the WT run. Nevertheless, a clear underestimation of the wet season (especially by the WT run) constrains the model performance at simulating streamflow.
The chapter concludes with a discussion on future efforts to better represent the model streamflow and therefore strengthen the line of research of this thesis; improvement of the deeper soil-groundwater linkage, improvement of surface runoff triggered by heavy rainstorms or snowcover depletion, and improvement of subgrid scale geological features representation.
In Chapter 5, a study of the groundwater effects on the land-atmosphere system and their seasonal and spatial variability is carried out, using the WT and FD simulations. Soil moisture shows that the water table makes the soil wetter over shallow water table regions, introducing a new pattern in soil moisture fields different to the climatic conditions and soil texture patterns. Analysing seasonal means, the water table influence on soil moisture is found to reduce partly seasonality, since soil wetness increase over the shallow water table regions is greater at water scarcity seasons (24.4% in spring, when evapotranspiration -ET -reaches the maximum rates, and 23.8% in summer, when soil moisture is at a minimum due to the lack of precipitation and high ET rates), and lower in autumn and winter (∼ 20%), when the surface water balance (P − ET, precipitation minus evaporation) is higher. The Long-Term Recharge is presented as the mean yearly water table flux, showing a positive pattern (upward recharge via capillary flux), as a result of lateral groundwater convergence and river infiltration, over extended shallow water table regions and river valleys. Groundwater recharge seasonal variability is found to respond to the surface precipitation and ET with two different cycles: 1) the negative recharge (downward drainage) is stronger during winter and spring, responding to drainage of infiltrated precipitation from the wet season, and weaker during summer and autumn after the dry season; 2) the positive recharge is strong in spring and maximal in summer (ET season in the Iberian Peninsula), due to the gradual decrease of the surface water balance, then during autumn it weakens significantly due to xiv the decrease of ET demands, and finally it reaches the minimum in winter, when precipitation presents the highest rates and ET the lowest.
The soil moisture increase by the water table presence via upward capillary fluxes translate to the atmosphere as ET enhancement. The ET difference between the WT and FD simulations present the spatial pattern induced in soil moisture by the water table (reaching significant differences over river valleys and shallow water table regions in the semiarid interior Iberia). The enhanced ET averaged over the Iberian Peninsula is 0.54 mm dy −1 in summer, when precipitation is the lowest and ET demands are high, 0.24 mm dy −1 in spring, when ET demands are the highest, 0.14 mm dy −1 in autumn and 0.05 mm dy −1 in winter, when the ET demands are very low. To conclude the chapter, convective precipitation enhancement and summer temperature decrease are discussed as potential effects of these water table dynamics results on fully coupled groundwater-land-atmosphere modeling research.
The choice of a 10-year simulation period including dry and wet episodes allows a deeper look into the water table influence on soil moisture, and in Chapter 6 the water table seasonal and interannual persistence and soil moisture memory that it induces is studied over the peninsula. The time evolution of the water table does not respond immediately to seasonal or yearly peaks and lows in precipitation, but it responds to long-term climatic conditions with 1-2 years delay. When the water table enters the soil-atmosphere system in the WT simulation, its persistence passes to the soil as soil moisture memory, and the soil "remembers" past dry and wet conditions, buffering drought effects on soil moisture and delaying the recovery after the drought in spite of wetter conditions.
The soil moisture memory induced by water table persistence is stronger over shallow water table regions, where soil and groundwater are hydraulically connected.
Analysing hydrological year (September-August) anomalies for precipitation, soil moisture and water table depth over a well known shallow water table region in the upper Guadiana basin (La Mancha Húmeda), the water table influence on soil moisture memory is inferred. Moreover, soil moisture memory is passed to the atmosphere as ET memory over this shallow water table region.
In Chapter 7, the groundwater effects on soil moisture and land-atmosphere fluxes, analysed in previous chapters at seasonal and interannual timescales, are studied as area average over the main Iberian river basins, as the basins integrate the groundwater influence through lateral flow convergence. This allows to investigate the groundwater role in the land-atmosphere system under different climatic conditions. At basin scale, the water table presence, through slow drainage and soil moisture accumulation during the wet season and upward capillary rise during the dry season, induces ET enhancement during spring and summer in the WT simulation. This groundwater influence on ET is more significant in southern basins of semiarid climates of water limited ET (yearly ET increases 28.4% in the Mediterranean Segura basin and 21.4% in the Atlantic Guadalquivir basin), but as the ET is xv less water limited towards the northern basins, the groundwater is less influential on ET fluxes (13.3% in the northern Miño basin, where ET is energy limited).
The time evolution of ET increase (WT-FD) follows the long-term trends of the td evolution, caused by consecutive dry and wet years. Hence, the water table persistence is passed to the atmosphere as ET memory at basin scale, and the yearly ET enhancement in the WT run depends not on only the previous wet season but also on the long-term td evolution. This is observed clearly over the Atlantic basins.
Over the Mediterranean basins, that present moderate wet season and variable dry season, the ET memory is constrained by the intensity of the dry season (less memory when the summer precipitation is especially low because the ET enhancement is high, and when the dry season is especially high because the ET enhancement is low, in both cases the intensity of the dry season takes over the td trend).
Finally, in Chapter 8, a summary of the main conclusions drawn through the investigation is presented, where an attempt to answer the initial Questions is produced.
A review of future efforts in this line of research concludes the thesis.
Early stage results of this investigation have been published in the article "The role of groundwater on the Iberian climate, precipitation regime and land-atmosphere interactions" [47]. However, the main results are in preparation to be submitted for publication in the article "Groundwater influence on soil moisture memory and landatmosphere interactions over the Iberian Peninsula" [26]. A high resolution estimation for the Iberian Peninsula Equilibrium Water Table Depth, from the a long-term recharge estimation obtained from LEAFHYDRO simulations is also in preparation to complete and submit the article "Water Table Depth and Potential Capillary Flux to the Land Surface in the Iberian Peninsula" [48]. xvi
La hipótesis de trabajo de esta tesis es que las Preguntas 2, 3 y 4 tienen SI como respuesta. Probar la hipótesis y buscar una respuesta a la Pregunta 1 son las motivaciones para el desarrollo de esta investigación en el Grupo de Física No Lineal de la Universidad de Santiago de Compostela.
El aumento de humedad de suelo inducido por la presencia de la capa freática a través de recarga positiva, repercute en la atmósfera en forma de aumento de ET.
La elección de una simulación de 10 años incluyendo periodos secos y húmedos permite un estudio a largo plazo de la influencia de las aguas subterráneas en la xxi Introduction

Natural Water Cycle
Water on Earth never stops; it is continuously and cyclically moving and transferring between different reservoirs in the atmosphere, land and sea; and it is continuously changing states between liquid, gas and ice. In space, these processes take place at planetary, continental, regional and local scales. In time, they occur at timescales of seconds, days, seasons, years and over millions of years. All these movements and processes are energetically driven by the sun and the Earth gravitation, and constitute the natural water cycle.
The main reservoirs and processes of the natural water cycle are illustrated in Figure   1. 1. The basic processes of the cycle are: evapotranspiration, precipitation, infiltration, recharge and runoff. Evapotranspiration is the combination of two fluxes; evaporation of water at the surface of the sea, rivers, lakes and land, and transpiration from living beings, especially plants, at the land surface.
The evapotranspirated water passes to the atmosphere as water vapour, then it is transported by atmospheric circulation and, under appropriate circumstances, it can condensate into liquid or solid water and fall back to the surface as precipitation. Precipitated water over the oceans or surface water reservoirs can reinitialize the cycle being evaporated back to the atmosphere, and precipitated water over land can be transported by gravitational forces towards rivers as surface runoff or infiltrate into the soil. Infiltrated water become then part of the soil moisture and can be evapotranspirated back to the atmosphere or percolate through the soil and into the groundwater reservoir as groundwater recharge.
The groundwater reservoir is the volume of water contained in the pores, fissures and fractures of the water-saturated strata of the Earth's crust. The water table is the surface that limits the saturated and non-saturated zones of the soil. Groundwater near the water table plays an important role in the natural water cycle through   different interactions (see Figure 1.1):   It can move up to the soil via capillary rise, or collect soil moisture drainage   provoking a water table rise. It can be transported as groundwater flow within the saturated zone, discharging ultimately to the oceans.
It can receive lake and river infiltration (seepage), or feed the rivers as part of the baseflow.
The water cycle is not a single mechanical circuit. It is necessary to understand and evaluate it a as a compound of different interconnected processes and cycles at different scales in time and space. The spatial scales for these interconnected circuits is continental, like the come and go of water evaporating from an ocean and falling back as rain, regional and local, like the circuits of water in river basins that characterize the region giving it a distinctive biological composition. Water participates in the different circuits and also can be transferred from one to another, like for instance evaporated water from the Pacific Ocean can be transported eastward by atmospheric dynamics, fall as rain on the Rocky Mountains, be transported eastward through the continent by the Mississippi River and end up in the Atlantic Ocean.
The residence times of water in the different reservoirs vary very much. For example, the period of renewal is 9700 years for polar ice, 2500 years for ocean water, 17 years for lake water, 1 year for soil moisture, 8 days for atmospheric moisture or only hours for biological water [107]. The time variability of the natural water cycle is introduced by its dependence on the atmospheric circulation and weather factors.
Such dependence on atmospheric dynamics makes it very important to understand the behaviour, variability and connection with the climatic elements of the water cycle reservoirs in climate modeling studies.

Climate variability in the Iberian Peninsula
The climate in the Iberian Peninsula is defined by its geographical situation (35N-40N, 10W-4E), its continentality and the stratospherical circumpolar vortex dynamics. The stratospherical circumpolar vortex is a large scale closed cyclical circulation around the Arctic Pole whose seasonal, annual or decadal changes in form and intensity influence the climate and weather in the Northern Hemisphere midlatitudes, and therefore the Iberian climatology is linked to the circumpolar vortex dynamics.
Due to the Iberian Peninsula location at the gates of Europe, very different types of air masses can reach it: cold and wet northern maritime Arctic (mA) masses from the Greenland Sea; cold and wet northern or northwestern maritime Polar (mP) masses from the Islandic region; cold and dry continental Polar or Arctic (cP, cA) masses from Escandinavia or the north of Russia; warmer but still wet sourthwestern maritime Tropical (mT) masses from the Azores High; warm and dry southern or southeastern continental Tropical (cT) masses from the north of Africa. This variability of air masses that can reach the Iberian Peninsula and the continentality that the orography gives to most of the peninsula result in a high climate variability, as the peninsula become a meeting point for different air masses.
Such climate variability is illustrated using the Koppen-Geiger climate classification.
This classification system uses series of monthly mean air temperature and precipitation, and defines distinct types of climate based on ranges of these values and their influence on the distribution of vegetation and human activity [38]. The map in

Precipitation variability in the Iberian Peninsula
Precipitation variability in time and space is a revealing factor to understand climate variability in the Iberian Peninsula.
About the seasonal regime, summer is the drier season and winter is the wetter, but autumn and spring are also wet seasons depending on the region studied inside the peninsula, and there is no dry season over the northern mountain ranges (type Cfb climate, Figure 1 are needed to explain precipitation in the eastern coast (far from the Atlantic influence) and the eastern half of the Cantabrian coast [74].
In order to analyse the spatial and interannual precipitation variability, the publicly available analysis daily precipitation gridded dataset IB02 (∼ 20 km spatial resolu-    This makes it a very interesting period to be included in long-term coupled land-atmosphere modeling studies. 3. The plot averaging data for the whole domain appears much more jagged than the rest, not reflecting the real behaviour at located regions. This calls for caution when analysing hydrological processes in the Iberian Peninsula, since Iberian average data may lead to mistaken conclusions. There is not a commonly accepted definition for droughts [62], but depending on the system affected and the impacts timescale they can be classified into 4 categories [116]: meteorological, agricultural, hydrological and socioeconomic. The criteria to define a drought differ from one public administration to the next, depending on local experiences. They take into consideration different thresholds of annual precipitation registered, longer or shorter periods, relations between water demands and resources or similarities with historical droughts. However, since groundwater reservoirs are less affected by lack of precipitation than by river discharge, indicators on the state of groundwater reservoirs are not normally accounted for. In basins where the connection between the streamflow and the natural discharge from aquifers (baseflow) is strong, it can be observed how lack of precipitation periods of 1-3 years (meteorological droughts) are not translated into hydrological droughts when analysing volumes of water in rivers or wetlands connected to the aquifers. A different problem altogether appears when the groundwater reservoirs are used for irrigation of cultivated lands or town water supply due to bad public planifications.

Droughts in the Iberian Peninsula
The groundwater reservoirs role as a mitigator of drought effects is one of the factors studied in this work (Chapter 6).

The groundwater link to the land-atmosphere system
The groundwater interactions with the land-atmosphere system are factors to be taken into consideration when climate and ecosystem modeling studies are performed. Hence, the inclusion of a dynamic water table as the lower boundary condition of the soil column constitutes an important step to improve the land surface representation in climate models. There has been a growing interest in the scientific community to represent goundwater-land-atmospher interactions in recent years at different spatial and temporal scales. Some such efforts and the key aspects of the problem are summarized in this section.

The soil moisture link to the atmosphere
Soil moisture is a key player in climate. Evapotranspiration is highly water-limited, and hence the availability of soil moisture constrains evapotranspiration to the atmosphere [39]. Many efforts have been made in the community to address the climate models sensitivity to land processes and the importance of a consistent soil moisture initialization given the lack of measurements [95]. Using [83] performed a simulation over North America using an initial version of the coupled groundwater-land surface model LEAFHYDRO for the warm season of 1997, they found a strong spatial correlation between the distribution of shallow water table and wet soil and pointed out the double role of the groundwater reservoir, shifting from being primarily a sink during the wet spring to being primarily a source for the soil water above during the dry summer.

The groundwater link to streamflow
Streams and lakes are connected through two-way water fluxes: the discharge from groundwater to the streams in gaining streams where the water table is above the river bed (typical of humid regions), and the water flux infiltrating into the groundwater from losing streams where the water table is below the river bed (typical of arid regions) [44]. In gaining streams, the groundwater reservoir may maintain the stream baseflow during dry seasons. Losing streams feed the groundwater reservoir during wet seasons, contributing to the groundwater memory during the following dry seasons. Hence, at a given location that presents a water table fluctuating around the river bed depth, the local rivers may act as gaining and losing streams, depending on the season. A proper representation of such groundwater-streams connection makes a difference in hydrological and climate studies.
In order to consider this groundwater-streams connection and close continental hydrological cycles, river routing models are coupled with land surface models.

The importance of a correct representation of groundwater interactions in the Iberian Peninsula
This work is a contribution to the cited efforts studying the influence of groundwater dynamics on the water cycle and the land-atmosphere fluxes over the Iberian Peninsula during a 10-year period. The Iberian Peninsula is a region of special interest for a groundwater-land-atmosphere interactions study provided its climatic variability detailed in Section 1. 2. The groundwater's role is investigated in following chapters at 3 levels in the Iberian Peninsula: 1. Impact on soil moisture dynamics, spatial variability and long-term memory. 2. Effects on land-atmosphere fluxes through evapotranspiration to the atmosphere. 3. Direct impact on surface waters through the groundwater-streams connection.   water table recovered carried out mainly by modest individual farmers, often with little planning and control on the part of governmental water authorities [46]. According to the Spanish Ministry of Natural Environment, 7% of the national territory is irrigated, 67.7% of this irrigated land uses water that comes from surface runoff and 28.2% (942,244 ha) uses groundwater. Studies included in the report Libro Blanco del Agua (2010) [6] estimated that the fraction of natural flow that actually runs through the Iberian rivers after the anthropogenic extractions decreases from north to south, presenting significantly low values for the main southern rivers (52% for Guadiana river in Badajoz, 44% for Guadalquivir river in Alcalá del Río), as illustrated on Figure 1.5. Similarly, in the last decade there has been a growing interest in the community to include groundwater exchanges with the land-atmosphere system as explained in Section 1. 3 LEAFHYDRO is a modeled representation of energy and water storages and vertical exchanges in the land surface. It includes prognostic equations for soil temperature and moisture for multiple layers, vegetation temperature, surface water including dew and intercepted rainfall, snowcover mass and thermal energy for multiple temporary snow layers, temperature and water vapour mixing ratio of canopy air, water table depth and river streamflow. Exchange terms in these prognostic equations include turbulent exchange, heat conduction, water diffusion and percolation in the snowcover and soil, longwave and shortwave radiative transfer, transpiration, precipitation, runoff, groundwater recharge, groundwater-river exchange and groundwater lateral transport.
LEAFHYDRO can work as an offline land surface model forced by atmospheric forecasted, reanalysed or observed fields. However, its very conception as a SVAT scheme makes it a tool to work coupled with atmospheric models, such as RCM

RAMS or RCM WRF -collaborative work between the NCAR and the USC Non-Linear
Physics Group is ongoing to carry out fully coupled dynamical groundwater-landatmosphere climate and weather simulations -. In this coupling with atmospheric models, LEAFHYDRO offers the ability to represent fine-scale variations in surface characteristics, since it is computationally inexpensive in comparison to the representation of processes in atmospheric models and can work at a higher spatial resolution [9,115]. Characteristics such as vegetation type, terrain slope, land use, bodies of water, soil moisture or water table depth vary considerably over the typical horizontal scales used in RCMs and GCMs, hence the finer resolution in the SVAT scheme represent more realistically processes that occur at subgrid scale when using atmospheric models. Such processes are easily understood, like different vegetation responses to soil moisture or radiative fluxes, or accumulation of moisture in valleys leading to relative drying in neighbour higher zones due to a shallower water table in the valleys or a quick rainfall runoff from sloping areas.
The energy and water storages and exchanges that LEAFHYDRO resolves are represented schematically in Figure   In the next sections the main LEAFHYDRO formulation is described. Firstly, Section 2.1 summarizes the main processes resolved by the model as a SVAT scheme (some calculations have changed since the original LEAF [63,64] and LEAF-2 [115] formulations), and then Section 2.2 is focussed on the groundwater dynamics scheme implemented.

Land surface and soil layers formulation
In this section the methodology used in LEAFHYDRO for the resolved soil crust and the land surface is summarized, starting from the soil layers and following up through snow and temporary surface water, vegetation, bare ground, canopy air, incoming precipitation and radiation.
The methodology used by LEAFHYDRO is simple. First, incoming and outgoing water and heat fluxes in Figure  The parameterization to calculate moisture fluxes within the soil is based on a multilayer soil model described by Tremback & Kessler (1985) [110], which was itself a modification of the scheme described by Mahrer & Pileke (1977) [73] and McCumber & Pielke (1981) [79]. Thus, the water vertical flux between adjacent unsaturated soil layers F ss (kg m −2 s −1 ) combines gravitational drain and capillary flux, and is given by the Richards' Equation, is the density of liquid water, K η (m s −1 ) is the hydraulic conductivity at a given volumetric water content η, Ψ (m) is the soil capillary potential and z (m) is height. Parameters K η and Ψ depend on the water content and the pore-size index of the soil. To compute such parameters, the model follows the Clapp & Hornberger (1978) [24] formulation: where b is the soil pore-size index and subscript ƒ denotes quantity at saturation.
From a simple water mass balance between incoming and outgoing water fluxes within each layer, LEAFHYDRO prognosticates the soil water content η of the given layer.
The soil heat flux between layers F hss (J m −2 s −1 ) is then due to the temperature gradient and the internal energy carried with the moisture flux, and it is given by is the specific heat of liquid water and L  (J kg −1 ) is the latent heat of fusion. The soil thermal conductivity is calculated using the Johansen's method [58]: where λ ƒ is the thermal conductivity in saturated state, λ d is the thermal conductivity in dry state (its value depends on the soil type) and λ e is a function representing the influence of the degree of saturation on λ: λ e = 1 + 0.7og 10 (η/ η ƒ ). For frozen and unfrozen soils, λ ƒ is calculated as: where ƒ  is the liquid water mass fraction, λ  is the thermal conductivity of liquid Once the heat fluxes between layers have been calculated, an energy balance prognosticates internal energy Q s (J m −3 ) of moist soil for each soil layer, relative to a reference state of soil and completely frozen moisture both at 0 • C. It is defined by where T s C ( • C) is soil temperature, ƒ  and ƒ  are the ice and liquid water mass fractions relative to the total water mass in the soil, m s and W s are the mass of dry soil and water, respectively, in kilograms per cubic meter of total volume (including water, soil and air), C  , C  and C s (J kg −1 K −1 ) are the specific heats of ice, liquid, and dry soil particles, and L  (J kg −1 ) is the latent heat of fusion. The use of Q s provides more information than temperature, as it represents the energy associated not only with temperature but also with latent heat of fusion. Temperature and liquid versus ice fraction are diagnosed from Q s , with knowledge of W s and m s . An increase of the ice fraction reduces the water flux within the soil provided that the frozen soil has zero percolation.

Snow and temporary surface water formulation
Snowcover, snow melt, rainwater and temporary streams are included in the definition of temporary surface water. That is, water that had reached the ground surface as precipitation but has not yet percolated into the soil or run off to permanent water bodies, like oceans or rivers.
LEAFHYDRO divides the snowcover into vertical layers depending on the depth of snow present, up to 3 layers. For each snow layer, the mass W n (kg m −2 ) and internal energy Q n (J kg −1 ) are prognosticated via energy and mass balances.
In analogy with Equation 2.6, the internal energy of snowcover relative to a reference state of ice at 0 • C is defined by where T n C ( • C) is the snow layer temperature, ƒ  and ƒ  are the ice and liquid water mass fractions relative to the total mass of the layer, C  and C  (J kg −1 K −1 ) are the specific heats of ice and liquid water and L  (J kg −1 ) is the latent heat of fusion.
Q n values between zero and L  do imply a mixture of ice and liquid at 0 • C in the snowcover, hence ƒ  is diagnosed as Q n / L  . The snowcover layer temperature T n is also diagnosed from Q n .
The water fluxes between snowcover layers F nn (kg m −2 s −1 ) and from the bottom snowcover layer to the soil F sn (kg m −2 s −1 ), and their associated heat fluxes F hnn (J m −2 s −1 ) and F hsn (J m −2 s −1 ) are calculated in the model by percolation of liquid water, as an excess over the layer liquid water holding capacity, and heat diffusion: First, the incoming fluxes from the atmosphere to the top snowcover layer are added to the layer mass and internal energy, and hence the fractional liquid water content is diagnosed from the internal energy.
Then, the liquid water exceeding 10% of the ice mass in the layer is percolated as F nn to the next snowcover layer.
To calculate the heat flux to the next layer LEAFHYDRO uses a downgradient diffusion relation following Adams & Brown (1973) [7], combination of the sensible heat in the presence of a temperature gradient and the latent heat carried by the vapour flux from warmer to cooler layers -note that the vapour flux is neglected because it is typically weak [7] -. The combined diffusion coefficient where T n (K) is the snow layer temperature, ρ n (kg m −3 ) is the density of the snowcover layer, and the last factor ƒ ρ n is a ρ n dependent function given by 3 . Additional heat flux carried by the percolating water must be taken into account, hence the net heat flux F hnn is given by where Q n is the internal energy of the upper snowcover layer carried with the moisture flux.
The mass and energy values are then updated in the layer and the process is repeated for the layer below.
For the lowest snowcover layer, the percolating liquid water into the soil F sn is limited by the top soil layer capacity to accept surface water before reaching saturation, and water exceeding this limit flows surface runoff. The net heat flux F hsn is calculated as After the snowcover fluxes have been calculated and mass and energy layer values have been updated, the temporary surface water parameterization follows and adjustment procedure described by Walko et al. (2000) [115] to keep the total snowcover layers number and individual snowcover layer thickness within prescribed bounds. This adjustment assures that no snowcover layer is to become too thin for stable numerical computation and, in order to well resolve the snowcover, the total number of snowcover layers is to increase when mass is added to snowcover if allowed by numerical computation stability up to 3 snowcover layers. Only if the snowcover contains some ice, multiple snowcover layers will be used. If so little snowcover exists that prognosis of its internal energy by explicit time differencing may be computationally unstable (in which case snowcover resides in a single layer), an implicit computation is done instead in which the snowcover and top soil layer are brought into thermal equilibrium.

Surface vegetation formulation
When there is vegetation on the surface and it is not cover by snow, the water and heat exchanges between vegetation and the surrounding canopy air parameterization is based on Avissar et al. (1985) [11].
The specific humidity at the leaf-surface interface q c (kg kg −1 ) is calculated as q c = r c q ƒ + r st q c r c + r st , (2.11) where q ƒ (kg kg −1 ) is the saturated specific humidity at the vegetation surface and q c (kg kg −1 ) is the specific humidity of the canopy air surrounding vegetation. The vegetation-canopy resistance r c (s m −1 ) is calculated as a function of the vegetation leaf area index (LA) and the friction velocity  * by whereas the stomatal resistance r st (s m −1 ) is defined as where r st mn is the minimal stomatal resistance for a given vegetation type and occurs when the stomata are completely opened. Each ƒ  function quantifies the influence of a specific environmental factor upon the resistance (R stands for solar global radiation, T for leaf temperature, V for vapour pressure difference between leaf and the surrounding air, C for ambient atmospheric carbon dioxide concentration and Ψ for soil water potential in the root zone), and is given by , (2.14) where the subscript  refers to the environmental factor, X  is the intensity of the factor , X b is the value of X  at ƒ  = 0.5, and S  is the slope of the curve at this point [12]. This stomatal resistance r st is limited by a variable lower threshold that corresponds to the maximum possible transpiration given the present atmospherical conditions.
The heat flux from vegetation to canopy air F hc (J m −2 s −1 ) is calculated from the temperature difference and the resistance r c , as where LA is again the dimensionless vegetation leaf area index, C p (J kg −1 K −1 ) is the specific heat capacity of air, ρ  (kg m −3 ) is the density of the air, and T c and T  (K) are the canopy air and vegetation temperatures, respectively.
Similarly, the water flux between vegetation and canopy air F c (kg m −2 s −1 ) is calculated from the humidity difference as where q c (kg kg −1 ) is the canopy air specific humidity, q ƒ (kg kg −1 ) is the vegetation specific humidity at saturation. σ is a factor of the exchanged water that comes from the vegetated surface, depending on the available water at the leaf surface, where   is the vegetated surface water content.   has an upper threshold of 0.22LAƒ  due to the shedding of moisture excess, and therefore σ is limited between 0 and ƒ 2/ 3  , where ƒ  is the vegetation fractional coverage.
The moisture exchange that transpirates through vegetation is given by The transpiration is taken from the moistest level in the root zone. Transpiration is limited between zero, when the canopy air is moister than the vegetation specific humidity at saturation, and the water available in the root zone.

Surface bare ground formulation
For the ground-canopy fluxes calculation, the formulation that LEAFHYDRO uses has been adopted from CLM (Community Land Model), which is the SVAT scheme developed by the NCAR Earth System Laboratory.
The formulation, similar to the described canopy-vegetation methodology, makes use of aerodynamic resistances to sensible heat transfer ,r h , and to water vapour transfer, r  , between the soil and the surrounding air, given by where θ (K) is potential temperature, q (kg kg −1 ) is specific humidity, the subscripts  and g stand for atmosphere and ground surface, respectively,  * (m s −1 ) is the friction velocity, θ * (K) is temperature scale and q * (kg kg −1 ) is humidity scale. The profile terms (θ  − θ g )/ θ * and (q  − q g )/ q * are calculated using the Monin-Obukov similarity theory [121].
The heat flux between the soil surface and the surrounding air F hsc (J m −2 s −1 ) is then calculated as , (2.20) where T g (K) is the temperature at the soil/snow surface (taken as the top soil/snow layer temperature).
The water flux between the soil surface and the surrounding air F sc (kg m −2 s −1 ) is given by where q sƒ e (kg kg −1 ) is the effective saturation specific humidity of the soil surface and r sc (m s −1 ) is the soil-canopy water flux resistance.
Following Lee & Pielke (1992) [55], q sƒ e is computed from the specific humidity of the soil surface at saturation q sƒ (kg kg −1 ) and the specific humidity of the canopy air q c (kg kg −1 ), as where g (m s −2 ) is acceleration due to gravity, Ψ g (m) is the soil water potential at the ground surface, R  (J kg −1 K −1 ) is the gas constant for water vapour and β is a surface wetness function, given by The ground dew/frost formation F de (kg m −2 s −1 ) is based on the saturation value of specific humidity at the soil surface temperature q sƒ , as (2.25)

Canopy-atmosphere fluxes
LEAFHYDRO calculates the sensible heat H (J m −2 s −1 ) and evapotranspiration ET (kg m −2 s −1 ) to the atmosphere as: Here, the temperature scale θ * , humidity scale q * and friction velocity  * are computed from the surface layer similarity theory as Louis et al. (1981) [70,115], using the canopy air values of humidity and temperature (calculated from surface balances as detailed further on in Section 2. 1.8) and atmospheric values of humidity, temperature and wind. The atmospheric variables may come from an atmospheric model if LEAFHYDRO works as the land surface scheme coupled to an atmospheric model, or from the forcing data variables if LEAFHYDRO works offline (see Section 3.3 for a description of the atmospheric forcing).
Note that the H and E fluxes are referred to in Figure 2.1 as F hc and F c , respectively. The surface water then may form temporary surface snowcover layers, when the atmospheric conditions allow it, or percolate into the soil if the top soil layer water content has not reached the soil capacity. If after percolation there is still precipitated water on the surface, it flows as surface runoff.

Radiative fluxes
As explained for precipitation fluxes, longwave and shortwave radiation fluxes that reach the surface are introduced into the LEAFHYDRO system either from the atmospheric model coupled to it or as external forcing when working offline.
Longwave radiation is emitted to the atmosphere or to the next reaching surface, absorbed, and reflected by vegetation, snowcover, soil and permanent water bodies. Snowcover, even when shallow, acts nearly as a blackbody to longwave ra- where α  , α n and α s are vegetation albedo, net albedo from all snow layers and soil albedo, respectively,  is again the non-covered by snow vegetation fractional coverage, τ n is net transmissivity of all snow layers and ƒ  is the fraction of total radiation absorbed by snow that is absorbed by layer .
The profile ƒ  is evaluated from the profile of transmissivities τ n  of each snowcover layer, which is a function of many factors including snow layer depth, density, snow grain size, and liquid water content. For simplicity, LEAFHYDRO parameterizes the snow layer  transmissivity as τ n  = exp(−εD n  ), (2.37) where D n  (m) is snowcover depth of layer  and ε (m −1 ) is an extinction coefficient.

Surface balances
After all surface water and heat fluxes have been calculated as explained in previous sections, LEAFHYDRO prognosticates surface values of temperature and humidity from heat and water mass balances.
The vegetation temperature T  is prognosticated from the heat balance at the vegetation surface between incoming and outgoing radiative, heat and moisture fluxes, And similarly, the water mass balance in the canopy air that LEAFHYDRO carries out to prognosticate the canopy air specific humidity q c (kg kg −1 ), is given by

Dynamic groundwater scheme
The LEAFHYDRO dynamic groundwater scheme was first presented by Miguez-Macho ]. The concepts that the scheme is based on are the dynamical behaviour of the groundwater reservoir and its interactions with the land-atmosphere system. There are three main interactions represented by the model:  Hence, the mass balance of the dynamic groundwater reservoir in a LEAFHYDRO cell is given by

Groundwater recharge formulation
The range of water table depths in extended regions with significant topographic and climatic variabilities (as in the case of the Iberian Peninsula) is very wide.
Hence, a numerical column soil-groundwater model has to find a balance between the need to resolve the upper most dynamic portion of the soil column and the con- where Ψ (m) is soil capillary potential and z (m) is height, evaluated in the layers referred to by subscripts 1 and 2. Applying the relationship between Ψ and η in Then, assuming even distribution of total soil water in the layer, the layer 1 soil water content η 1 that the model calculated in the soil fluxes routine as described in Section 2.1.1, can be expressed as And finally, knowing the water  The water content of point B is initially obtained by linear interpolation between A and C (water content in the virtual layer containing C is part of the model initialization). Given water content at A and B, the flux between the two can be calculated.
In the same manner, an auxiliary layer is added below the water table, containing point D and with equal thickness as the layer containing C. The water content gradient between C and D (layer containing D is saturated) determines the flux between the two, which is the groundwater recharge R. Knowing the fluxes above and below, the new water content η C of the layer containing C can be determined by mass balance. The change in water content in the virtual layer is added to or taken away from the groundwater reservoir as Δt, calculated similarly to Equation 2.46 from the groundwater recharge as is the saturation soil water content for the soil at the water table position depth.

Groundwater lateral flow formulation
The methodology for the calculation of the lateral flow within the saturated groundwater reservoir carried out in LEAFHYDRO was discussed by The flow transmisivity T is calculated as where K L ƒ (m s −1 ) is the lateral hydraulic conductivity at saturation and z (m) is vertical downwards direction.
Given the lack of observations for this lateral hydraulic conductivity at saturation, the model uses the anisotropy ratio α, that relates K L ƒ with the hydraulic vertical conductivity at saturation K V ƒ as α = K L ƒ / K V ƒ . The values taken for the anisotropy ratio α are well within the range observed in nature and based on the clay content of the soil [42]. But available data for vertical hydraulic conductivity only covers the top soil metres, hence continental and regional scale groundwater modeling needs to make assumptions on the vertical distribution of hydraulic conductivity. Exponential decrease with depth is commonly assumed for hydrological model over the scales of metres, Decharme et al. (2006) [32] found K V ƒ to improve its performance using an exponential profile for discharge simulations over the Rhône Basin with the LSM ISBA (Interaction Soil-Biosphere-Atmosphere). LEAFHYDRO uses available data for K V ƒ up to 1.5 m deep and assumes exponential decrease downwards, in the form  Under scenario B, the water table appears below 1.5 m deep, thus T is given by The e-folding depth ƒ (m) reflects the sediment-bedrock profile at a location. It is a complex function of climate, geology and biota, but the balance depends strongly on terrain slope. As a first order approximation to capture features of the water The ƒ depth value is limited to 4 m when β ≥ 0. 118.
In order to ensure that flow from the considered cell to the neighbour is the same as from the neighbour to the considered cell under the same hydraulic potential, LEAFHYDRO calculates T for both cells involved and uses the average of the two.

Groundwater-streams exchange formulation: Gaining and losing streams
For the calculation of the groundwater-streams flux, LEAFHYDRO distinguishes between gaining streams, when the water table is above the riverbed elevation and feeds the streams, and losing streams, when the water table is below the river elevation and the streams drain water towards the groundwater reservoir. The sketch in Figure 2  Combining the first to parenthesis in the right side of Equation 2 [83] and is justified by the dynamics of the groundwaterstreams contact area [29,30].
The equilibrium part of the river conductance, ERC, represents the hydraulic connection between the streams in a cell and the groundwater reservoir in the longterm river channel evolution. Hence, ERC is linked to the equilibrium water

River flow scheme
LEAFHYDRO closes the water cycle resolving the streamflow discharging through the river channels in the model domain and ultimately to the ocean. The streamflow q (m 3 s −1 ) is calculated as where S s (m 3 ) is the surface water storage and K s (s) is the residence time for the surface water in the cell.
The surface water storage is calculated from a mass balance every river scheme timestep, as   LEAFHYDRO assumes rectangular channel cross section, thereforeĀ h is given bȳ  Table) simulation.
In addition, to help isolate the role of the groundwater, another simulation is performed with the groundwater scheme switched off. This second simulation will be referred to hereafter as FD (Free-Drain) simulation.
The FD simulation uses the commonly adopted free-drain approach, where soil water is allowed to drain out of the soil column at a rate set by the hydraulic conductivity at the water content of the bottom soil layer. The potential drawback of this approach is that the escaped water is no longer available for subsequent dry period evapotranspiration. It should work very well where the water table is deep and the soil is sandy, but where the water table is shallow and the soil is clay-rich, it may underestimate the soil water storage and overlook water persistence. This may be one of the reasons that recent climate reanalysis must rely on significant soil water nudging where water is added to or removed from the soil column to meet atmospheric demands [94].    [56], and made publicly available. For the LEAFHYDRO 2.5 km horizontal resolution, the elevation data is aggregated. Figure   3.1 shows the resultant model domain topography used in the LEAFHYDRO simulations.

Vegetation parameters
The vegetation type or land cover field for the work domain is shown in Figure 3

Soil parameters
The soil textural classes used in LEAFHYDRO are defined by the USDA (United States Department of Agriculture) as fractions of silt, clay and sand 1 Table 3.1: Soil parameter for the USDA textural classes used in LEAFHYDRO. Parameters are, from left to right: volumetric soil water content at saturation η ƒ , soil capillarity potential at saturation Ψ ƒ , conductivity at saturation K ƒ , pore-size index b and volumetric specific heat of dry soil C d .

Equilibrium Water Table Depth
The climatic equilibrium water As initial condition for this work, the climatic EWTD (Equilibrium Water Table Depth) for the Iberian Peninsula in Figure 3.3 is used. This field is a first guess to introduce and understand patterns in the peninsula water table depth distribution.  The result is the Iberian Peninsula EWTD in Figure 3.5, used as initial condition in the WT simulation. Comparing this EWTD to the initial EWTD (

Soil moisture initialization
For the soil moisture LEAFHYDRO initialization, the soil water content is calculated in preprocessing from bottom to top, parting from the knowledge of the equilibrium water

River parameters
The model requires an estimation of the riverbed mean elevation in the cellz r in order to calculate the equilibrium river conductance and the groundwater-streams flux in gaining streams (Section 2.  The flow of the main high resolution stream within every low resolution cell is then followed, highlighting the stream (red streams in Figure 3.6).
The distance made by this high resolution main stream is taken as the low resolution river length L.
The low resolution river slope s r is taken as the average slope for all high resolution cells that take part in the main high resolution stream, where the high resolution slopes have been previously calculated from the flow direction ƒ d and the elevation dem.
The low resolution drainage area A d is calculated aggregating the area of all high resolution cells within a low resolution cell, and then accumulating it from all cells addressed to a given cell with the use of the low resolution ƒ d.
Finally, the width  r is calculated using the recharge R and the drainage area where Q m is the annual mean discharge passing through a river section, approximated for this calculation by the accumulation of flow Q = RA d for the cells along the low resolution stream.

Atmospheric forcings
The atmospheric forcing data fields at the surface, needed for the LEAFHYDRO simulations, were extracted from the ECMWF ERA-Interim 3 reanalysis database [2,4,5].
A processing scheme has been introduced into the LEAFHYDRO forcing routines However, given the importance for the water cycle of the incoming water in the system through precipitation, the regional higher resolution analysis of daily precipitation dataset over Spain and Portugal (IB02) introduced in Section 1.  Some of these models consider the baseflow in a manner largely uncorrelated to the process it is intended to quantify, like for example making it solely dependent on the moisture content of the lowest soil layer [119] or the water table depth [88,117].
Even when in the model, through calibration or any other strategy, the baseflow amount is accurate and the water

Water table depth and time evolution: validation with observations
Actual water table depth (td) observational data is used in this section to validate the model performance in terms of capturing the water table position and its time evolution.
Time series graphs in

Observational td data
The observational td data were provided by the IGME (Institute of Geology and

Mining of Spain), several Confederaciones Hidrológicas (Spanish agencies managing the main watersheds within the country) and the SNIRH (National Information
System for Hydrological Resources of Portugal).

Model performance
Some studies incorporating explicitly groundwater dynamics in land surface modeling find the water table effects to be negligible at depths below 5 m [66,77], but the potential contribution by upward capillary flux to evapotranspiration of water

Modeled streamflow: validation with streamflow observations
In this section the LEAFHYDRO streamflow that results from the simulated water cycle in the WT and FD simulations is analysed and the model is validated over the Iberian Peninsula using observational streamflow data.
Streamflow is a suitable parameter to validate land surface-hydrology models, since streamflow time cycles result from the models interactions between the different reservoirs represented, and there is access to observational data. However, it is important to mention that the goal of LEAFHYDRO is not to simulate the surface water flow and use these results for hydrological purposes, but to carry out simulations of the complete water cycle inland and the land-atmosphere interactions.

River flow scheme in the FD simulation
The river flow scheme used by LEAFHYDRO was described in Section 2. 3 where Q h (m 3 s −1 ) is the hillslope overland runoff given from the temporary surface With this approach the water budget in the FD simulation is closed as it is in the WT simulation, but the water draining through the bottom model soil layer goes directly to the rivers, and therefore it is no longer available for the soil or land-atmosphere interactions.

Observational streamflow data
A monthly streamflow observation dataset has been compiled for validation. The   The disconnection of the observed streamflow with the natural streamflow in the Guadalquivir basin responds to anthropogenic influence.

Validation and WT-FD streamflow comparison
To illustrate the behaviour of both the WT and FD simulated streamflow in compar-  From the graphs on the right; the rivers dry out in summer practically every year in the FD run, whereas in the WT run the groundwater reservoir feeds the rivers during the dry season using accumulated wet season recharge, and keeping the summer baseflow higher and closer to observations. This is clearly observed in the Miño, Tajo and Ebro stations. 3. The seasonal cycle is differently represented by the WT and FD simulations: The FD simulation reaches higher and closer to observations streamflow values in the winter months.
As the dry season starts, the rivers get drier below the observed values in the FD run, while the WT run maintains higher summer baseflow.
After summer, the WT simulation follows the observation trends, recovering streamflow with the autumn precipitation from September/October.
However, in the FD simulation it takes longer to recover from the dry sum- E is defined as where T is the number of times in the time series, Q obs is the observed streamflow and Q mod is the modeled streamflow. In general hydrology model simulation,   Top: Correlation coefficient r for the WT simulation monthly streamflow (dark blue), correlation coefficient rmm for the WT simulation monthly mean streamflow (light blue), correlation coefficient r for the FD simulation monthly streamflow (red), correlation coefficient rmm for the FD simulation monthly mean streamflow (orange), Nash-Sutcliffe coefficient E for the WT simulation monthly streamflow (green) and Nash-Sutcliffe coefficient E for the FD simulation monthly streamflow (yellow). Bottom: Annual mean streamflow (Hm 3 yr −1 ) in the WT simulation (blue), FD simulation (red) and observed (green).

Winter streamflow underestimation
The clear underestimation of the flow during the wet season in the WT simulation observed in Figure 4.4 is attributed to two different factors: the precipitation input and the model itself. 1. The first factor is a low input precipitation in high mountainous areas. The density of observations used to build the analysis precipitation dataset, IB02 [16,54], that forces the LEAFHYDRO simulations correspond to that of the AEMET and IPMA monitoring network, and after the quality selection criteria the mean distance between pairs of stations was 7 km in Spain and 11.7 km in Portugal, with 28 km and 41.6 km as maximum distance, respectively. Such density might not be enough to capture the high winter precipitation peaks over elevated zones or very wet areas, where heavy rainstorms trigger surface runoff.   At some locations the soil changes to less permeable rocks or even the groundwater might be trapped in confined aquifers, in which case the infiltrated water does not reach as deep as the water table and is available for the soil quicker. At other locations, the porosity might increase due to rock crevices in the deeper soil, in which case the connection between the groundwater and the land surface is quicker than the model representation.
The deeper soil-groundwater linking representation needs to be improved in LEAFHYDRO, but it is out of the scope of this work. In any case, the results in following chapters based on the groundwater-land surface connection might be conservative, since a better representation of the deeper soil hydraulic properties is expected to increase the groundwater linkage to the land surface as suggested by the winter streamflow underestimation.
Subgrid scale surface geological features, like rocks less permeable than the grid cell soil type representation, also may strengthen these problems. A better representation of subgrid scale surface runoff processes and deep soil properties is considered for future work in this line of research.

Discussion
This chapter objective is to validate the 10-year LEAFHYDRO WT simulation with water table depth and streamflow observational data, and hence support the thesis findings in following chapters . The water table pattern

Water table control on soil moisture
The experiment in this work consists of long-term simulations using the LEAFHY-

Water table pattern.
In addition to the patterns of soil texture and climatic conditions, the WT simulation soil moisture plots (third column) present an overall wetter soil and a more defined soil moisture spatial structure, that responds to the shallow water table depth patterns (fourth column). The seasonal variability in the td plots is not as strong as the seasonal differences in the precipitation and FD soil moisture plots, and hence the water table presents significant seasonal persistence that is translated to the WT soil moisture plots.
The seasonal soil moisture differences between the WT and FD simulations account  In relative terms (blue columns), the water table greater influence on soil moisture at water scarcity seasons (24.4% wetness increase in spring and 23.9% in summer) is appreciated, reducing partly soil moisture seasonality. Note that the data represented in Figure 5.3 are ∼ 3.2 times the given data averaged for the whole Iberian Peninsula.
To summarize, the soil moisture fields are not only a direct response to the soil water holding capacity and the seasonal climatic different conditions; the water influence is greater at water scarcity seasons; in spring, when evapotranspiration reaches the maximum rates, and summer, when the soil moisture is at a minimum due to the lack of precipitation and high evapotranspiration rates.

Water flux across the water table: Recharge
The introduction of the groundwater interactions with the unsaturated soil as the model lower boundary condition changes the nature of the water flux across the water table or recharge. The study of this variable gives an understanding of the connection between the groundwater reservoir and the soil.
The recharge is the water flux directly responsible for the soil moisture differences between the WT and FD simulations analysed in Section 5. 1. The connection between the unsaturaded soil and the groundwater reservoir through the recharge flux is bimodal, depending on the soil wetness conditions: Negative recharge appears when the groundwater reservoir acts as a sink to infiltrated water that comes from precipitation and exceeds evapotranspiration to the atmosphere at the surface.
Positive recharge appears when the groundwater reservoir takes the role of water source to soil moisture via upward capillary flux as atmosphere evapotranspiration demands exceed the water input from precipitation.
In this section, the Long-Term Recharge obtained as the yearly mean water table flux along the WT simulation is presented and analysed first, and the recharge seasonal variability is studied later.

Long-Term Recharge
In methods commonly used, groundwater recharge is inferred by tracer studies from streams flow (limited to river basins spatial scale) [10] or indirectly diagnosed from other variables, as R = P − ET − Q h , where the evapotranspiration ET and surface runoff Q h are calculated using LSMs without groundwater dynamics, and precipitation P usually comes from analysis datasets [47].
In this work, the methodology is different, the water table flux for a large region as the Iberian Peninsula is explicitly calculated every timestep (900 s) in the WT simulation. Hence, the Long-Term Recharge, understood as the yearly mean water table flux along the WT 10-year simulation, has been obtained considering explicitly long-term feedbacks between the groundwater reservoir, soil moisture, evapotranspiration and stream baseflow, as the result of a long-term land surface and hydrology model simulation that incorporates groundwater dynamics. The Long-Term Recharge is presented in Figure 5.4 (left), together with the mean yearly precipitation forcing (right). In

Seasonal variability
The groundwater recharge presents a strong seasonal variability. Figure   A clear seasonal evolution in the recharge can be appreciated from Figure 5.5 plots: The winter plot is mostly red and resembles the structure of the autumn and winter precipitation plots in Figure 5.2, therefore the groundwater reservoir is a sink to the wet season precipitation water that infiltrates through the soil layers.
In spring the winter rainfall keeps flowing down through the soil layers and to the groundwater reservoir, maintaining the negative recharge pattern as strong as it was in winter, but spring precipitation is not as high and the evapotranspiration demands start to rise, so there begin to appear important blue upward flux patterns over valleys and shallow water table regions.
The summer plot shows a maximum in recharge all over the peninsula. The precipitation input is minimal during summer, weakening the red negative pattern, and the evapotranspiration demands from the atmosphere keep forcing the groundwater reservoir to act as a water source, strengthening the blue upward patterns that already appeared in spring.
In the autumn plot the role of the groundwater reservoir as a water source starts to weaken since autumn precipitation is enough to cover autumn evapotranspitation demands, and drainage through the water table is not as strong as it was in spring, even though the precipitation input is similar or even stronger for the wetter regions in autumn ( Figure 5.2), since the infiltrated water takes time to reach deeper water tables.
In conclusion, the groundwater recharge presents a clear seasonal cycle following the precipitation and evapotranspiration cycles. Downward recharge is strong during winter and spring, responding to drainage of infiltrated precipitation from the wet season, and weaker during summer and autumn after the dry season. Upward flux is appreciated over shallow water table regions; it presents significant values in spring, when the evapotranspiration season starts, and reach the maximum in summer when the surface balance is minimal (P − ET = −0.80 mm dy −1 ), then during autumn it decreases significantly due to the decrease of evapotranspiration demands, and finally it reaches the minimum in winter when precipitation presents the highest rates and evapotranspiration the lowest.

Enhanced Evapotranspiration (ET)
The fluxes across the water table result in water table depth

Discussion
The objective of this chapter is to study the seasonal and spatial variability of the water The negative recharge is stronger during winter and spring, responding to drainage of infiltrated precipitation from the wet season, and weaker during summer and autumn after the dry season.
The upward flux is strong in spring, due to the beginning of the evapotranspiration season, and maximal in summer, when the surface balance is minimal.
Then, during autumn the upward flux decreases significantly due to the de-  The Iberian Peninsula presents, as noted from the precipitation plots in Figure 5

The effects of water table persistence on soil moisture memory
The water table persistence and its connection to the soil through upward capillary fluxes are expected to induce certain memory of past atmospheric conditions into the soil. Over regions where the water table is shallow, the groundwater-soil link is more important as seen in Section 5.1, and the longer timescale of variation of the water table position affects soil moisture.

Analysis over the Iberian Peninsula
The plots in Figure  -On the eastern coast, the hy1 positive precipitation anomaly is especially high, and the water table appears shallower than the mean during the next 3 years, even though the precipitation anomalies in the region keep changing.    precipitation anomalies do not explain soil moisture anomalies in the WT simulation as strongly as in the FD simulation (left).
The orange line in Figure 6.4 (right) represents the spatial correlation between the td anomalies for the past 2 years (hy t , hy t−1 ) and soil moisture anomalies at hy t , whereas the dark yellow line represents the spatial correlation between the td anomalies for the past 3 years (hy t , hy t−1 , hy t−2 ) and soil moisture anomalies at hy t . Their high value and their upper trend during the dry middle years and the last 2 years can be explained as soil moisture memory of water table positions during former years while precipitation follows a decreasing trend (hy2-hy4, hy7-hy9).

A closer look: La Mancha Húmeda
La Mancha Húmeda is a Biosphere Reserve situated in the upper Guadiana Basin. It comprises the Tablas de Daimiel National Park, which corresponds to the core area, the Alcázar Lake and the Lagunas de Ruidera National Park. The Equilibrium Water Table is found very shallow at this region (see Section 3.2.2), and therefore it is a marked wet spot that helps understand the water table influence on soil moisture memory at a finer scale.

Soil moisture memory induced over the localized shallow water table zone
In Figure 6.5 a series of plots represent a 250 × 225 km 2 area containing La Mancha Húmeda. The zoomed area is highlighted with a green contour in Figure 6.2 (middle row, first plot). The plots show, chronologically for the 9 hydrological years simulated; total yearly precipitation anomalies (top row, simple zoom of top row in Figure 6.2), soil moisture anomalies in the FD simulation (second row), soil moisture anomalies in the WT simulation (third row), and end of hydrological year td anomalies (bottom row, simple zoom of the middle row in Figure 6.2).
From Figure 6.5, the following is noted: Soil moisture anomalies in the FD simulation (second row) are a clear response to precipitation anomalies (top row). The averaged soil moisture anomalies (colour bars below second and third row) change when considering the water table influence. For instance, the mean soil moisture anomaly over the zoomed area in hy7 results positive due to positive precipitation anomaly, whereas in the WT simulation it is negative due to water table persistence.
From Figure 6.5 it is concluded that soil moisture patterns in the FD simulation follow directly the evolution of precipitation patterns, whereas in the WT simulation soil moisture patterns reflect a combination of precipitation and td patterns. The FD approach, without groundwater representation, lacks the memory induced in the WT simulation by the groundwater implementation, which is very apparent over shallow water table areas. This memory is bimodal, since the water table persistence makes the soil "remember" dry and wet periods, and therefore drought effects are buffered but also last longer.

The effects of soil moisture memory on evapotranspiration (ET)
In Section 5.3, it was discussed an enhancement of ET fluxes to the atmosphere in the WT simulation . It was more significant over shallow water table regions and water scarcity periods, when analysing the seasonal means over the Iberian Peninsula. But, how will the increased soil moisture memory in the WT simulation affect land-atmosphere fluxes evolution in time?
The next set of plots in Figure 6.6 attempts to answer this question over the zoomed region, and focussing on summer ET, when the effects are clearer. The plots show, chronologically for the 9 hydrological years simulated; total yearly precipitation anomalies (top row), summer daily ET anomalies in the FD simulation (middle row), and summer daily ET anomalies in the WT simulation (bottom row).
From Figure 6.6, the following is pointed out: Summer ET anomalies in the FD simulation (middle row) reflect clearly the precipitation anomalies pattern for each year (top row), as do the soil moisture anomalies in the FD simulation in Figure 6. 5.
In the WT simulation, however, a new pattern of enhanced ET areas is super-   Figure 6.2 (middle row, first plot). Each column represents a whole hydrological year (hy1 to hy9) simulated. Rows from top to bottom: total yearly precipitation anomalies (mm), summer ET anomalies (mm dy −1 ) in the FD simulation and summer ET anomalies (mm dy −1 ) in the WT simulation.

Discussion
Soil moisture memory refers to dry or wet anomalies of longer persistence in time, as compared to atmosphere conditions. The objective of this chapter is to study the water

Main Iberian basins
The identification of the basins is performed using a routing scheme that follows the main rivers and all their tributaries upwards from the mouth to the sources. For this scheme the river parameters A d (m 2 ; accumulated drainage area) and ƒ d (flow direction), calculated as detailed on Section 3. 2.4, are used.

Seasonal and long-term time evolution
The seasonal, annual and long-term behaviour of the groundwater-soil-atmosphere system is studied next over the main Iberian rivers basins (Figure 7.1). The td series (red lines in left graphs) presents a very clear annual cycle in all basins, following precipitation (blue bars in left graphs) and ET; upward trend during autumn and winter, since the surface balance   Groundwater makes always the soil wetter in the WT run, as discussed in Chapters 5 and 6.
The difference in soil moisture presents lows during precipitation peaks because the soil is wet in both the FD and WT runs, and hence the water table presence is not as influential.
On top of the latter lows, there is a clear annual cycle in the soil moisture difference; increasing gradually from autumn towards spring since the water table presence slows down drainage during the rain season in the WT run, and then decreasing drastically in late spring or early summer. 3. The latter drastic decrease in soil moisture difference is explained by the annual peak of enhanced ET in the WT run (WT-FD ET difference; turquoise bars at the bottom in left graphs); In spring ET demands are maximal and therefore ET enhancement reaches a peak; a) in the WT run, the water accumulated in the top soil during the rainy season due to slow drainage, plus the dominance of the upward capillarity flux later during the dry season, cover ET demands, whereas b) in the FD run, the top soil accumulated less water during the rainy season and there is no upward capillarity flux, resulting is less water availability to cover ET demands.
The timing of the enhanced ET is different for the different basins, depending on the soil water availability from precipitation. In the Miño basin the period of intense enhanced ET is short, only during summer. As water availability decreases towards the central and southern basins due to less intense and shorter rainy season, the ET enhancement period starts earlier, in early spring or even late winter (like during the central years of simulation in the Guadiana and Guadalquivir basins), and lasts longer if the dry season has been especially dry (like in 1995).
The amount of total yearly enhanced ET is similar for all Atlantic basins.
However, the enhanced ET is more significant in relative terms as the basins are drier towards the south (see Figure 7.4), ranging from 13.3% in the Miño basin to 21.4% in the Guadalquivir basin.
Note that data in Figure 7.4 is averaged for the whole basins (not only   shallow water table cells) ET is mostly energy limited in the Miño basin. In the rest of the basins ET is water limited, therefore the groundwater influence on ET enhancement over the shallow water table regions is stronger.  The total enhanced ET yearly evolution in Figure 7.5 gives a measure of this memory for the five Atlantic basins, reflecting the long-term td trends. Hence, the ET enhancement during the dry season depends not only on the previous wet season precipitation but also on the water table position.

Mediterranean basins
A different precipitation regime is observed for the Mediterranean basins ( Figure   7.1); moderate with precipitation all year in the Ebro basin, and moderate wet season with a marked dry summer in the Júcar and Segura basins. The total rainfall amount decreases significantly towards the south. The td annual cycle and the reach of its peaks and lows follow the rain and ET seasons as described for the Atlantic basins, but the intensity of the cycle is lower due to the lower accumulated precipitation during the wet season.
The severity of the drought (growing from north to south) is as intense as seen in the southern Atlantic basins, but its effects on the long-term water table evolution are intensified by the moderate precipitation during the last 3 years, which stops the depression of the water table during the drought but it is not enough to rise the water table position to shallower depths as seen in the Atlantic basins. 2. The annual cycle of the WT-FD soil moisture difference is lees intense than in the Atlantic basins because the wet season is more moderate. In the Ebro basins where summer is nearly as wet as winter, the cycle is harder to detect. 3. About ET enhancement in the WT run; The annual cycle of the WT-FD ET is the same as seen for the Atlantic basins. However, the differences between basins are not so strong; the reach of the maximum ET enhancement is very similar every year and the intense enhancement period starts slightly earlier in the basins with less intense rainy season.
As observed for the Atlantic basins, the enhanced ET is more significant in relative terms as the basins are drier towards the south: 14.1% in the Ebro basin, 20.3% in the Júcar basin and 28.4% in the Segura basin. 4. About soil moisture memory; The power spectrum (right graphs in Figure 7.6) shows the increase of soil moisture memory as the WT simulation present higher amplitudes for longer periods. The memory grows from north to south. The soil moisture difference presents a long-term deepening trend during the drought and then a steady position during the last 3 years, translating the water table persistence into soil moisture memory, since the water table never recovers initial depths after the drought.
The total enhanced ET yearly evolution in Figure 7.7, however, do not reflect the td and soil moisture difference long-term trends as clearly as seen in Figure

Discussion
The strength of the groundwater-soil coupling depends on the water table depth, but the effects of the coupling differ under different climatic conditions. In this chap- During the wet season (autumn-winter) the water table rises due to drainage of precipitation, but its presence (WT run) slows down the drainage as compared to the FD run, and therefore the soil moisture difference between both runs also follows an upward trend. Then, when the ET season begins in spring, the soil moisture difference is maximal and there is more soil water availability to meet ET demands in the WT run, causing a marked peak in the ET difference (WT-FD; ET enhancement) evolution. During spring and summer the higher soil water availability in the WT run continues due to capillary rise and there is ET enhancement until the next wet season, when the cycle starts again. The ET enhancement is more significant in the drier southern basins, where ET is more water limited (21.4% ET enhancement in the Atlantic Guadalquivir basin and 28.4% in the Mediterranean Segura basin). In the northern Miño basin where ET is not so much water limited as energy limited, the ET enhancement is less significant (13.3%).
The 10-year period of simulation presents a long drought from 1990 to 1995/1996 in all basins except the Miño basin. The water table follows the long-term precipitation trend with a slow td depression along the drought, and then gradually recovering to shallower a position during the last 3 years when precipitation is high. This long-term td trend is passed to the soil moisture difference time evolution, and therefore the soil moisture potential contribution to ET "remembers" past dry and wet years. In the Atlantic basins, the soil moisture memory induces ET memory, more clearly in the southern drier basins, where the water table influence on the land-atmosphere fluxes depends not only on the previous wet season intensity but also on the long-term td evolution. In the Mediterranean basins, this ET memory is constrained by the intensity of the dry season, since during years of especially dry summer ET enhancement increases above the td trend, and during years of especially wet summer ET enhancement decreases below the td trend.

Chapter 8 Conclusions
Land-atmosphere coupling is a key factor in climate modeling research. In this context, it is essential for the scientific community to achieve a correct representation of the processes and storages that affect land-atmosphere interactions in GCM and in summer, when ET demands are lower than in spring but the precipitation is at a minimum, moderate in autumn, when precipitation is enough to cover ET demands, and minimal in winter, the season of highest precipitation and lowest ET demands. 4. The groundwater-soil coupling over shallow water Analysing yearly soil moisture anomalies over the 10-year period simulated, the new pattern of soil moisture evolution reflecting the water table pattern is captured as induced soil moisture memory. This soil moisture memory is bimodal; 1) soil moisture "remembers" past wet conditions through the water table persistence at shallow positions, buffering drought effects, and 2) soil moisture "remembers" past dry conditions through the water table persistence at deep positions, causing a delay in drought recovery. 7. The soil moisture memory is passed to the atmosphere as ET enhancement memory in the WT run. Analysing the integrated water table effects on the land-atmosphere system and their long-term evolution over the Iberian river basins, the td long-term evolution is reflected by the soil moisture difference (WT-FD) evolution, or in other words, the water table persistence is reflected by the long-term evolution of the soil moisture potential contribution to ET enhancement in the WT run. Hence, ET enhancement "remembers" past dry and wet years. This ET memory is strong in the Atlantic basins, but in the Mediterranean basins it is constrained by the intensity of the dry season; in years of especially dry summer ET enhancement increases above the td trend, and in years of no dry season (summer nearly as wet as winter) ET enhancement decreases below the td trend.

Future work
The line of research followed in the USC Non Linear Physics Group during recent years, producing initial conditions for the water table depth in the Iberian Peninsula [47], the Amazon [41], or even global patterns [40], in order to later evaluate the groundwater influence on LSM simulations as done in this thesis, has a very clear Work is ongoing to elaborate a high resolution Equilibrium Water Table Depth for the Iberian Peninsula, which will reproduce the spatial distribution of the water table in the peninsula more realistically than the one used as initialization in this thesis [47], since the atmospheric influence in the form of climatic groundwater recharge will be a long-term estimation from LEAFHYDRO several decades simulation.