1. Introduction
Arid and semiarid regions of the southwest United States and northwest Mexico are influenced significantly by the North American monsoon (NAM). This atmospheric circulation pattern is responsible for generating a large proportion of the annual rainfall, ranging from 40% to 80% (e.g., Douglas et al. 1993; Vivoni et al. 2008) and up to 85% of the annual streamflow during July, August, and September (Gochis et al. 2006). The in-phase relationship between precipitation and radiation suggests that correspondence may exist between runoff mechanisms, which typically depend on seasonal wetness, and the land–atmosphere interactions arising during the NAM. Several studies have investigated how soil moisture plays a role in land–atmosphere exchanges and their impact on subsequent rainfall generation in the NAM region (e.g., Small 2001; Xu et al. 2004; Vivoni et al. 2009). This interaction arises from a positive soil moisture–rainfall feedback (Eltahir 1998) that has been shown to operate in the region during the summer season, aided by ecosystem greening (e.g., Dominguez et al. 2008; Méndez-Barroso and Vivoni 2010). To our knowledge, however, an analysis has not been performed of the relationship between evapotranspiration (ET) and the underlying runoff mechanisms in the NAM region.
Various studies have investigated the partitioning of precipitation into ET and runoff and their relationship with soil moisture using climate models (e.g., Luo et al. 2007; Hamlet et al. 2007). For example, Hamlet et al. (2007) found that trends of ET and runoff ratios in the western United States were strongly linked to the trends in seasonal precipitation. Other studies have related the aridity index to runoff ratios and compared empirical equations to climate model simulations with the purpose of exploring the relationships between ET and runoff (e.g., Arora 2002; Koster and Suarez 1999; Sankarasubramanian and Vogel 2002). Prior analyses, however, have not taken into account the interplay between seasonal responses of runoff mechanisms and evapotranspiration. Given the heterogeneous landscape in the NAM region (e.g., Mascaro and Vivoni 2010), the spatial variability in runoff components and their link to land surface fluxes merits more detailed attention. Therefore, the objective of this study is to explore if a relation exists between seasonal runoff mechanisms and ET within the NAM region.
Prior studies have also addressed streamflow seasonality in the NAM region (Brito-Castillo et al. 2003; Gochis et al. 2006). However, understanding the spatiotemporal variability of streamflow remains an elusive challenge for several reasons. First, most regional river basins, particularly within Mexico, remain ungauged (or poorly gauged) because of inadequate rainfall, weather, and/or streamflow records. Sparse data limit the ability to observe hydrologic variables of interest at the appropriate spatiotemporal scales for streamflow forecasting. Second, the NAM region experiences large temporal variations of meteorological characteristics (Gochis et al. 2007; Gebremichael et al. 2007; Nesbitt et al. 2008) and in ecosystem dynamics (Vivoni et al. 2007b; Forzieri et al. 2011). Meteorological conditions also occur upon a template of high topographic variability (Coblentz and Riitters 2004), which lead to large spatial gradients in landscape conditions that are difficult to capture within streamflow simulations. Finally, the region is characterized by pulsed flood events occurring in a summer season of high wetness concurrent with elevated evapotranspiration demands (Vivoni et al. 2010). These phenomena lead to a complex interplay between event-scale and seasonal responses that have not been captured in prior analyses or forecasts (e.g., Gochis et al. 2006; Muñoz-Arriola et al. 2009).
A few studies have investigated streamflow generation and runoff processes in NAM areas within Mexico (García-Oliva et al. 1995; Gutiérrez and Hernandez 1996; Mora and Iverson 1998; Descroix et al. 2002a,b). Descroix et al. (2007) evaluated the role of soil and vegetation characteristics on infiltration- and saturation-excess runoff mechanisms at four sites in Mexico through a modeling approach and comparisons with field data. The authors found that both runoff mechanisms were scale dependent and decreased with larger catchment areas, which is consistent with work by Goodrich et al. (1997) in Arizona. More importantly, Descroix et al. (2007) argued that specific combinations of vegetation and soil types influenced the partitioning between infiltration- and saturation-excess runoff. Catchments with forests, woody savannas, or grasslands on permeable soils were found to be dominated by saturation-excess runoff, whereas areas with less dense vegetation or pastures on lower conductivity soils exhibited higher infiltration-excess runoff. These field observations indicate that a rich set of runoff mechanisms are possible and that numerical models could be useful for to capturing their spatial organization.
In this study, we use a distributed hydrologic model, the Triangulated Irregular Network (TIN)–based Real-time Integrated Basin Simulator (tRIBS; Ivanov et al. 2004; Vivoni et al. 2007a), to explore the spatial and temporal variability and patterns of streamflow, runoff mechanisms, and evapotranspiration in a large river basin (~9350 km2) within the NAM region. Physically based, distributed hydrologic models are tools for generating streamflow predictions that explicitly represent landscape properties and meteorological forcing and their associated spatiotemporal changes (Smith et al. 2004; Wood et al. 2011). Because of the sparse nature of meteorological data in the NAM region, we explored the use of the North American Land Data Assimilation System (NLDAS; Mitchell et al. 2004) as forcing for the spatially distributed model. NLDAS fields were compared, and in some cases adjusted, with ground observations in the study basin prior to the model application, as suggested by Luo et al. (2003) and Pinker et al. (2003). Driving hydrologic models with spatially and temporally variable meteorological fields is considered critical in the NAM region because of the pulsed nature of storm events and the strong gradients in atmospheric conditions imposed by complex terrain (Vivoni et al. 2009).
Through the model application, we pursue the question of whether a link exists between runoff generation and surface fluxes in the heterogeneous landscapes of the study basin. Since the model captures gradients in terrain, soil, vegetation, and aquifer properties, it provides streamflow forecasts that are linked to underlying runoff mechanisms, including infiltration-excess, saturation-excess, and groundwater exfiltration types (Vivoni et al. 2007a). Thus, we determine whether runoff mechanism patterns could be useful for inferring ET (or vice versa). We examine the impact of different meteorological forcings by comparing fields from a sparse network of ground stations, raw products from NLDAS, and the NLDAS products adjusted using available data. These scenarios are evaluated with respect to their impact on the spatiotemporal links between runoff generation and ET. Identifying the value of NLDAS products in the NAM region would have important implications for other river basins with sparse data—in particular within northwest Mexico. Combining modeling frameworks that capture spatially explicit meteorological and watershed conditions is also important to generate useful predictions for water managers and decision makers in the region (e.g., Robles-Morua et al. 2012).
2. Methods
a. Study region
The study site is the upper Sonora River basin (USR), the largest of three subbasins that form the Sonora River, covering 44% of the basin. Figure 1a shows the location of USR basin. The basin has an arid to semiarid climate with a mean annual rainfall from 355 mm near Hermosillo to 711 mm at higher elevations (Hallack-Alegria and Watkins 2007). The headwaters are located near the mining town of Cananea, Sonora, and are shared with the transboundary San Pedro River (e.g., Serrat-Capdevila et al. 2007). The USR flows south until its final destination in the Adolfo Félix Valdez reservoir, which is known as “El Molinito” (Fig. 1b). Elevations range from 228 to 2603 m (mean elevation of 1033 m), slopes range from 0° to 69° (mean slope of 10.8°), and aspects are primarily east and west facing (Fig. 1c), as determined from a 28-m digital elevation model (DEM) from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). Basin physiographic features include rugged bedrock slopes, dissected foothills and piedmonts, alluvial valleys along the river, and lower-elevation flatlands (INEGI 2000a).
(a) USR basin location in Sonora, Mexico; (b) subbasin delineation, stream network, and ground station locations; (c) 28-m ASTER elevation distribution; (d) 463-m MODIS vegetation classification; (e) 446-m INEGI–INIFAP soil classification.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
The distribution of vegetation and soils in the region is strongly linked to the complex topography (Coblentz and Riitters 2004; Vivoni et al. 2007b). A 463-m vegetation classification from the Moderate Resolution Imaging Spectroradiometer (MODIS) was used to describe land cover in the basin (Fig. 1d). The dominant vegetation types in the USR basin are open and closed shrublands (85% of the total area) and include desert and subtropical grass, shrub, tree, and succulent species (Vivoni et al. 2007b, 2010). Other important vegetation types include woody savanna and conifer forests at higher elevations and riparian woodlands and croplands. Brown (1994) and Paredes-Aguilar et al. (2000) provide detailed descriptions of the vegetation communities in this region. Soil distributions are known from soil surveys (resampled to 446 m) and include phaeozems, fluviosols, lithosols, luviosols, regosols, xerosols, and yermosols (INEGI–INIFAP 2001; Fig. 1e). Lithosols and regosols occupy 67% of the basin. Lithosols with shallow depths are located in steep hillsides, while regosols are located near alluvial areas but have poor organic contents. According to Descroix et al. (2007), soils in arid and semiarid areas of northern Mexico have characteristics that generally promote infiltration-excess runoff.
The ground observations consist of 13 weather and rain gauge stations from three different networks (Fig. 1b; Table 1): Comisión Estatal del Agua (CEA), Comisión Nacional del Agua (CONAGUA), and Arizona State University and Universidad de Sonora (ASU–UNISON). Stations were selected because of their full year of continuous data from 1 June 2007 to 31 May 2008, identified as the period with highest data availability from 2004 to 2008. CEA and ASU–UNISON sites contain hourly rainfall data, with additional weather parameters at CEA stations. CONAGUA sites with daily rainfall observations are used here to improve spatial coverage. Using the daily CONAGUA stations, we determined that the study period had average rainfall as compared to the long-term mean. For example, the average annual rainfall (50 years) at the cnaban station (Table 1) was 444 mm as compared to the 450 mm recorded for the study period.
Characteristics of the weather station and rain gauge sites in the USR basin available for the study period (1 Jun 2007–31 May 2008), including annual rainfall estimates from three meteorological forcing scenarios. The coordinate system is Universal Transverse Mercator (UTM), zone 12N, 1984 World Geodetic System (WGS84) datum.
A manual stream gauging location (“El Oregano”) is located upstream of the El Molinito reservoir. While this site was used to delineate the USR boundary (Fig. 1b), the daily stream discharge measurements at the site are subject to high uncertainties because of the discharge calculation methodology and the impact of several impoundments and diversions upstream of the gauging site. Further, as detailed in the following, the hydrologic model focuses on simulations of the tributary inflows to the main river stem in the USR basin rather than the integrated discharges in the main stem, reducing the need to use the daily flows at the basin outlet.
b. Distributed hydrologic model application
The tRIBS model operates on individual catchments represented by a TIN consisting of elevation, channel, and boundary nodes that capture topographic features and allow a reduction in the number of computational elements (Vivoni et al. 2004). A TIN is associated with a Voronoi polygon network that serves as the finite-volume domain for mass balance and flux calculations. For the USR basin, we selected 291 subbasins draining into the USR main river based on having a Horton–Strahler order greater than 3 at the confluence with the main stem. Figure 1b shows the boundaries for all subbasins, which occupy 92% of the USR basin area, derived from terrain processing of the 28-m ASTER DEM. For each subbasin, we generated a TIN based on the slope criteria method described by Vivoni et al. (2004). This method is based on the topographic relevance of elevation points in describing a domain and was applied to guarantee that 30% or more of the original DEM cells were preserved, as suggested by Vivoni et al. (2005). The subbasins had a mean value of the horizontal point density (d = nt/ng, where nt and ng are the number of TIN and DEM nodes, respectively) of d = 0.49 and ranged from d = 0.31 to d = 0.66.
Individual tRIBS simulations were conducted in a serial fashion on a high-performance computing platform. Since the parallelization capabilities in tRIBS (Vivoni et al. 2011) were not available yet, the large domain was divided into individual subbasins (a similar strategy is followed in the parallel version). For each subbasin model domain, we processed terrain, soil, vegetation, bedrock depth, and initial groundwater distributions to provide distributed model parameters and the boundary and initial conditions. Because of the large number of subbasins, the spatially explicit basin properties were derived for the entire USR basin from either remotely sensed or local data sources and then extracted for each particular subbasin. Vegetation types from MODIS were based on the International Geosphere–Biosphere Programme (IGBP) classification consisting of 17 land cover classes. Soil texture types at a resolution of 1:250 000 from INEGI–INIFAP (2001) were based on the Food and Agriculture Organization (FAO) classification. For each subbasin, vegetation and soil classes were directly mapped to the model domain to populate parameters for individual Voronoi polygons.
The model parameterization was carried out by specifying soil hydraulic and thermal properties, vegetation characteristics, and channel routing parameters. In this work, we guided our model parameterization by the calibration and testing study of Vivoni et al. (2010), where field measurements and numerical modeling results were compared at an eddy covariance tower and in a ~100-km2 basin located in the Sonora River (outside of the USR). Soil parameter values were based on applying pedotransfer functions (Schapp et al. 2001; Acutis and Donatelli 2003) to the data obtained from the World Soil Database (FAO 1988; Batjes 2002). Vegetation parameters were based on Vivoni et al. (2010) and literature estimates gathered by NLDAS (http://ldas.gsfc.nasa.gov/nldas/) for different ecosystem types. Table 2 lists the parameters used in our study for the major soil and vegetation parameters, including the percentage of basin area occupied by each class. Given the number of subbasin simulations and the lack of hydrologic data in the region, we limited the study to using a consistent set of model parameters derived from prior work and literature values. As shown in Table 2 and Fig. 1, spatial variations in model parameters and their magnitudes are expected to be sufficiently high to induce patterns in the basin response that can lead to different runoff mechanisms within individual subbasins.
Model parameters for major soil types and vegetation classes in the USR basin. The percentages of area of each soil and vegetation class relative to the total USR basin are shown.
Initial and boundary conditions were obtained from studies conducted by CEA. A water table distribution was derived by interpolating manual groundwater depths in wells sampled in 2007. The depth to water table is used in the model as an initial condition for the soil moisture profile under the assumption of hydrostatic equilibrium (Ivanov et al. 2004). Initial groundwater conditions were allowed to reach dynamic equilibrium by performing a 3-yr continuous spinup period using periodic forcing. This time was determined by performing a series of preliminary simulations using up to 5 years of spinup time. It was found that 3 years of spinup time was sufficient, which is consistent with prior work (Vivoni et al. 2005). We then used a set of electrical soundings and data on aquifer boundaries determined by CEA (2005) to create a distribution of the depth to bedrock in the USR basin. In regions with no aquifer or electrical sounding data, we used a 1-m soil depth determined by Vivoni et al. (2010) from a series of soil profiles. The procedures for deriving the depth to bedrock and initial groundwater level allowed us to capture spatial variations throughout the USR basin that alter the interactions between the unsaturated and saturated zones and lead to different runoff mechanisms.
c. Numerical experiments and meteorological scenarios
The modeling experiments in the USR basin consist of three meteorological forcing scenarios (or cases) during a 1-yr simulation. Figure 2 describes the time-varying forcing and spatial inputs used in the modeling scenarios, called GAUGES, NLDAS RAW, and NLDAS ADJ. Time-varying spatial forcings were generated for the period 1 June 2007–31 May 2008 at an hourly resolution. The period was selected to start prior to the NAM season to ensure dry initial conditions that matched the available data from groundwater well records. The period selection also allows for storage changes to undergo an appropriate return to initial conditions during a year and thus could represent an alternative definition for a “water year” in the NAM region. This procedure was essential for ensuring that the periodic forcing applied for 3 years yielded a state of dynamic equilibrium in the groundwater system at the end of the spinup.
Schematic representation of three USR simulations (GAUGES, NLDAS RAW, and NLDAS ADJ) resulting from a combination of forcing from ground stations and NLDAS.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
The GAUGES simulation consists of meteorological forcing to the tRIBS simulations from ground stations, including rainfall (mm h−1), atmospheric pressure (hPa), air temperature (°C), relative humidity (%), and wind speed (m s−1). A Thiessen polygon interpolation was applied to distribute the point-scale measurements. For meteorological data other than rainfall, this implied the use of three weather stations (stations ceacan, ceabac, and ceaure in Table 1). For rainfall, 13 locations provided hourly rainfall data since the CEA and ASU–UNISON networks were augmented with daily CONAGUA rainfall estimates that were downscaled to hourly periods using the hourly fraction of daily rain at the nearest high-resolution gauge.
Because of the sparse nature of these ground observations, we also evaluated the application of the high-resolution (hourly, 
3. Results and discussion
a. Basin-scale variability of rainfall, streamflow, and water balance components
We evaluated the spatially distributed rainfall estimates from the three scenarios for the entire USR basin. Figure 3 shows the annual precipitation of the GAUGES, NLDAS RAW, and NLDAS ADJ for the study period. GAUGES leads to a spatially discontinuous rainfall map because of the Thiessen interpolation and the large variations among stations (Table 1; Fig. 3a). Note the higher rainfall amounts in the northwestern and southeastern areas of the basin. In addition, a distinction exists between drier areas along the central western mountains and wetter areas in the opposing eastern slopes. These general features are also captured in NLDAS RAW, though their spatial extents vary considerably (Fig. 3b). Clearly, the higher resolution of NLDAS fields (~12 km) significantly improves the spatial depiction of rainfall in the USR basin. As shown in Table 1, the corrections to NLDAS resulted in an increase in precipitation for NLDAS ADJ at most locations, while retaining higher spatial variability (Fig. 3c). Daily spatial adjustments improved the bias between GAUGES and NLDAS ADJ to B = 0.98 (from B = 0.86 for NLDAS RAW) and reduced the RMSE between collocated sites from 117.9 mm yr−1 (NLDAS RAW) to 79.5 mm yr−1 (NLDAS ADJ). Further, the basin-averaged rainfall was nearly matched after the adjustment, with GAUGES at 542 mm, NLDAS RAW at 430 mm, and NLDAS ADJ at 518 mm.
Spatial variability of annual precipitation for study period for (numbers in parenthesis represent basin-averaged rainfall amounts) (a) GAUGES (542 mm yr−1), (b) NLDAS RAW (430 mm yr−1), and (c) NLDAS ADJ (518 mm yr−1).
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Figure 4 presents the basin-averaged temporal dynamics of water balance components for each scenario, including mean areal precipitation (P), evapotranspiration (ET), surface runoff (R), and depth-averaged soil moisture in the 10-cm surface layer (θS) and top 1-m root zone (θR) at hourly resolution. As expected, temporal variation in the hydrologic system is bimodal with a stronger NAM season (76%–80% of the annual rainfall in July, August, and September) and a weaker winter period. Nevertheless, the time-varying fluxes and states indicate that large differences are present among the three forcing products. The GAUGES scenario exhibits higher P, ET, R, and soil moisture (θS and θR). We attribute this to the overestimation of rainfall in the northern subbasins due to the sparse sampling by a few stations (Fig. 3). It is interesting to note how differences in P translate to basin-averaged surface runoff, where baseflows are sustained in GAUGES and NLDAS ADJ, but not in NLDAS RAW. In addition, θS and θR are impacted by differences in precipitation amount and timing. For GAUGES and NLDAS ADJ, a carryover of θR is present from summer to winter seasons—an indication that the NAM can set antecedent wetness conditions for fall and winter storm events. This impact is observed in the lower runoff response of NLDAS RAW to a storm event in December. Clearly, the NLDAS RAW scenario has limited runoff and soil moisture dynamics (except in the surface layer) because of the relatively low rainfall. Evapotranspiration is also lower in the NLDAS RAW (ET = 250 mm) case, which is an indication of the stronger soil moisture limitation in relation to the other scenarios (ET = 467 and 417 mm yr−1 for GAUGES and NLDAS ADJ). The carryover θR also supports ET from plant transpiration during the fall, winter, and early spring in GAUGES and NLDAS ADJ.
Simulated basin-averaged precipitation (P, mm h−1), evapotranspiration (ET, mm h−1), runoff (R, mm h−1), soil moisture in top 10 cm (θS, m3 m−3), and soil moisture in the top 1 m (θR, m3 m−3) for (a) GAUGES, (b) NLDAS RAW, and (c) NLDAS ADJ.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
The areal-weighted dynamics in the USR basin arise from geographical variations in subbasin characteristics. Figure 5 displays the percent contribution to the total streamflow volume by subbasin. Note that weak correspondence with the total rainfall volume (Fig. 3) is due to the rainfall–runoff transformation. Both rainfall characteristics (intensity, duration, and spatial location) and landscape properties (soil, vegetation, and groundwater depth) influence this transformation. Streamflow in northern regions differed significantly among the scenarios. In northeastern areas, subbasins contributed 16% (GAUGES), 1% (NLDAS RAW), and 2% (NLDAS ADJ) of the total runoff, while northwestern regions contributed higher fractions of total runoff: 25% (GAUGES), 12% (NLDAS RAW), and 21% (NLDAS ADJ). Despite these differences, all scenarios preserved a major feature of the rainfall forcing in that western portions generated more streamflow than eastern ones. This result was caused by lower conductivity (KS) soils in western subbasins (~10 mm h−1), as compared to eastern areas (~16 mm h−1), and higher slopes in western mountains (Fig. 1c). Better agreement among products is found in southern regions because of higher station density. These subbasins contributed the largest percent of total streamflow with 51% (GAUGES), 77% (NLDAS RAW), and 72% (NLDAS ADJ). Figure 5d indicates how the annual volume is distributed during the year. Clearly, rainfall bimodality is translated into two streamflow periods, with higher summer amounts that peak in July (GAUGES) or September (NLDAS RAW and NLDAS ADJ cases) for this year. Note how NLDAS scenarios shift streamflow from summer to winter—an indication that the GAUGES case overestimated rainfall and runoff in summer period and underestimated these in the winter. A detailed analysis of the spatial variations of the dominant runoff mechanisms and their relationship to land features is explored in the following section.
Spatial distribution of the percent contribution of each subbasin to the total streamflow in the USR basin (numbers in parenthesis represent the total streamflow volumes): (a) GAUGES (574 Mm3 yr−1), (b) NLDAS RAW (85 Mm3 yr−1), and (c) NLDAS ADJ (254 Mm3 yr−1). (d) Monthly streamflow as a percentage of the total annual volume for the three scenarios.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
b. Runoff generation and underlying mechanisms at the subbasin scale
Spatial distributions in streamflow contributions are related to the underlying runoff mechanisms simulated in the model, as discussed in Vivoni et al. (2007a, 2009). Figure 6 presents the percentage of each major runoff mechanisms in the simulations (RI, RS, RG) relative to the total runoff (RT) determined at each subbasin outlet. Perched return flow (RP) was not generated because of the lack of vertical variations in KS in the model. As expected for arid and semiarid regions with high-intensity rainfall, RI dominates runoff generation for all scenarios (Fig. 6, top). It is also evident that RI was more prevalent in western subbasins, where lithosols had lower Ks. Nevertheless, there are subbasins in the northern, eastern, and southern areas with large runoff percentages from RS. These cases occur on more permeable regosols, where sufficient rainfall input surpasses the soil storage capacity. Interestingly, landscape conditions in eastern subbasins promoted higher amounts of RG because of the relatively shallow groundwater depths in this region (Fig. 6, bottom). For example, a total of 15 (GAUGES), 22 (NLDAS RAW), and 23 (NLDAS ADJ) subbasins with total areas of 600, 921, and 997 km2, respectively, had RG/RT of 100%. The underlying physical mechanisms for spatial differences in runoff generation are discussed in the following by selecting four subbasins, labeled in Fig. 5a (BCN01, BAC33, BAV48, and BAV79), that represent the range of spatial variability in runoff generation mechanisms. BCN01, BAC33, BAV48, and BAV79 are naming descriptors used to identify subbasins located in different parts of the USR.
Spatial distribution of runoff mechanisms as a percentage of total runoff (RT) for (a) GAUGES, (b) NLDAS RAW, and (c) NLDAS ADJ. (top) Infiltration-excess runoff RI, (middle) saturation-excess runoff R, and (bottom) groundwater exfiltration RG.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Topographic index distribution in four selected subbasins: (a) BCN01, (b) BAC33, (c) BAV48, and (d) BAV79. Also shown are boundaries for rain gauge Thiessen polygons and soil map units, where numbers represent the associated value of saturated hydraulic conductivity (Ks, mm h−1).
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Characteristics of four selected subbasins (BCN01, BAC33, BAV48, and BAVB79), including terrain, soil, vegetation, rainfall, evapotranspiration, and streamflow properties (see Fig. 5a for locations).
The runoff mechanisms responsible for streamflow generation in each subbasin are analyzed in Figs. 8 and 9 for the GAUGES and NLDAS ADJ scenarios (results omitted for NLDAS RAW). For each case, the percent of time of occurrence (%) of a runoff mechanism (RI, RS, or RG) is shown as a function of λ. This allows an assessment of the effect of terrain features on runoff production, as well as a measure of the most frequent mechanism. For both scenarios, BCN01 and BAV48 exhibit a dominant contribution of RI, while BAC33 and BAV79 have more significant periods of RS and RG, respectively (also see Fig. 6 for comparison). From these distributions and Fig. 7, we can identify how specific runoff mechanisms are related to subbasin locations and landscape characteristics. For example, subbasin BCN01 is characterized by low Ks, which leads to the dominance of RI, in particular for GAUGES (Fig. 8a), which had higher rainfall intensities and uniform forcing as compared to NLDAS ADJ (Fig. 9a). Similarly, BAV48 had RI as a major runoff mechanism because of soils with low Ks, but exhibited higher Rs, especially for regions with higher values of λ located near the channel network. For NLDAS ADJ, RI and RS are codominant (Fig. 9c), with RG contributions in hillslope hollows (intermediate values of λ) and along the main river (high values of λ). In contrast, subbasin BAC33 has higher Ks, which leads to a dominant role for RS when forced with sufficient rainfall, as in GAUGES (Fig. 8b; 599 mm). A clear increase in RS with higher λ indicates the role of low-lying floodplain areas in producing runoff (Vivoni et al. 2005). For NLDAS ADJ (Fig. 9b; 410 mm), the low rainfall reduced all runoff occurrences. Interestingly, BAV79 had the most significant RG under both scenarios, implying less sensitivity to rainfall forcing (Figs. 8d and 9d). As with RS, groundwater exfiltration increases in frequency within convergent areas (high values of λ), which is an indication that permeable soils and steep slopes sustain a shallow aquifer near the main channel.
Runoff mechanism occurrence (percent of total time) variation with the topographic index [λ = ln(Ac/tan β)] for four subbasins: (a) BCN01, (b) BAC33, (c) BAV48, and (d) BAV79, in the GAUGES scenario. Vertical bars represent standard deviations of the percent of time of runoff occurrence computed for the corresponding values in the λ bins (bin size = 2 units of λ).
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
As in Fig. 8, but for the NLDAS ADJ scenario.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Comparisons between these subbasins exemplify the complex interactions between rainfall characteristics (spatial distribution, intensity, and total amount) and landscape properties (soil conductivity, slope, and terrain index) that generate runoff from multiple mechanisms. BCN01 and BAV48 represent northern, western, and southern areas where RI is the major component and is controlled primarily by low Ks. For these sites, RS may become codominant under conditions of lower rainfall intensities in NLDAS ADJ, imparting a dependence on terrain features. BAC33, on the other hand, is representative of eastern and northern areas where terrain-mediated RS is the major component when sufficient rainfall occurs. Under insufficient rainfall amounts, permeable soils in BAC33 reduce runoff for all landscape positions. Finally, BAV79 is a good example of a small set of subbasins in eastern areas that produce RG consistently and where contributions from RI and RS occur under particular rainfall forcings.
c. Interactions between runoff mechanisms and evapotranspiration
Spatial variations in the composition of runoff mechanisms should be linked to the land–atmosphere interactions occurring in each subbasin. Figure 10 shows the basin-averaged water balance in the GAUGES and NLDAS ADJ cases, including precipitation (P), evapotranspiration (ET), channel runoff (R), and root zone soil moisture (θR). As expected, the subbasin water balance varies considerably from the spatially averaged estimates in the USR basin (Fig. 4). Furthermore, the differences in P, ET, R, and θR among individual subbasins are striking, and substantially larger than variations between GAUGES and NLDAS ADJ. Subbasins with RI as the dominant component (BCN01 and BAV48) exhibit large R pulses of short-duration, more rapid recessions in θR and shorter periods of sustained ET. In contrast, in subbasins where RS and RG codominate with lower RI (BAC33 and BAV79), R is sustained over longer periods as baseflow, high θR carries over between seasons, and ET is sustained throughout the year. Over the entire period, subbasins with dominant RI tend to have lower ET/P than those subbasins where RS and RG have a significant contribution, as highlighted in Table 3.
Water balance components for four subbasins (BCN01, BAC33, BAV48, and BAV79) under (a) GAUGES and (b) NLDAS ADJ scenarios, including precipitation (P, mm h−1), evapotranspiration (ET, mm h−1), runoff (R, mm h−1), and root zone soil moisture (θR, m3 m−3).
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
To further investigate subbasin differences, we inspect the daily relations between ET and surface soil moisture (θS) in Fig. 11. For clarity, these are shown as the piecewise linear regressions of the total ET and daily averaged θS obtained at the scale of each subbasin for the GAUGES and NLDAS ADJ scenarios. Piecewise regressions of the ET–θS relation are obtained using a nonlinear optimization algorithm (Vivoni et al. 2008) and describe how water limitations in soils influence losses to the atmosphere (Rodríguez-Iturbe and Porporato 2004). For each scenario, a distinction is made between winter [December–February (DJF)] and summer [June–September (JAS)] seasons to capture the variation of the atmospheric demand due to seasonality in radiation. As expected, daily ET at a given θS is higher in JAS. Differences among the subbasins are also more pronounced during the NAM when variations between soil moisture promote maximum ET values ranging from ~4 to 7 mm day−1. Subbasins dominated by RI (BCN01 and BAV48) have lower ET under dry soil conditions (θS < 0.1) in JAS for both scenarios (Figs. 11b,d). This is an indication that infiltration-excess runoff promotes low infiltration, leading to rapid drying in shallow soils, which cannot sustain high ET rates. A clearer distinction of the effect of RI on the ET–θS relation is observed for the GAUGES scenario, likely because of the more frequent runoff production in JAS (Fig. 10). In contrast, subbasins BAC33 and BAV79 with higher RS and RG exhibit higher values of ET across all θS (GAUGES) and for dry soils (NLDAS ADJ). This result points toward sustained ET even when shallow soils are dry, primarily because of transpiration from the deeper root zone, which remains wet for longer periods because of groundwater contributions. The ET–θS relation for subbasins with RS and RG also suggests that ET rates are stressed over shorter soil moisture ranges and reach higher maximum (unstressed) values (also see Vivoni et al. 2008).
Daily soil moisture–evapotranspiration relation for (a),(b) GAUGES and (c),(d) NLDAS ADJ scenarios during the (a),(c) winter (DJF) and (b),(d) summer (JAS) months. Total daily ET (mm day−1) and daily averaged surface soil moisture (θS) are spatially aggregated in each subbasin (BCN01, BAC33, BAV48, and BAV79). Lines represent piecewise regressions of the daily relationships.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
We carried out a similar analysis for the entire USR basin to corroborate if the subbasin relations between runoff mechanisms and ET were valid across the region. Figure 12 presents the annual and seasonal ET as a function of the fraction of infiltration-excess (RI/RT) and saturation-excess (RS/RT) runoff relative to the total runoff (RT) in all subbasins (291 total). As expected, summer (JAS; bottom row) contributions to the annual ET (top row) are larger than winter (DJF; middle row) values for the GAUGES and NLDAS ADJ cases. A discernible pattern emerges from this analysis. Subbasins with greater RI/RT exhibit a decrease in annual ET, whereas subbasins with higher RS/RT fractions show an increase in annual ET. While this is clearer in GAUGES, the pattern is also found for the NLDAS RAW (not shown) and NLDAS ADJ scenarios, suggesting that the pattern is robust with regard to the type of rainfall forcing. Subbasins with an increasing RI/RT tend to have less infiltration and support lower ET amounts. In contrast, subbasins with increasing RS/RT exhibit greater infiltration, which allows higher ET. In addition, subbasins with greater RS/RT are also subject to groundwater contributions (Fig. 6) that sustain high soil moisture and ET rates. We evaluated how increases in RI/RT were accompanied by decreases in RS/RT by estimating the linear correlation (R2 values) of RI/RT versus RS/RT for all subbasins. Results show a strong inverse relation for the annual and JAS scales as compared to DJF. In GAUGES, R2 values for annual and JAS were 0.80 and 0.88, while it is 0.09 for DJF.
Annual and seasonal relations between evapotranspiration (ET, mm) and the fraction of infiltration-excess runoff (RI/RT) and saturation-excess runoff (RS/RT) for (a) GAUGES and (b) NLDAS ADJ. (top) Annual, (middle) winter months (DJF), and (bottom) summer months (JAS). Here Ri represents either RI or RS. Dashed lines representing linear regressions between ET and Ri/RT are included for visualization purposes only.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Overall, the variations in Ri/RT can yield annual ET differences of ~400 mm (GAUGES) and ~150 mm (NLDAS ADJ). The seasonal relations also show that the annual trends between ET and runoff mechanisms are consistent for winter and summer. Seasonal patterns, however, are stronger in JAS for GAUGES and in DJF for NLDAS ADJ, which is consistent with the shift in monthly runoff toward winter (Fig. 5d). As a result, NLDAS ADJ summer season exhibits low sensitivity to the runoff mechanisms. Similarly, we conducted an analysis of the relation of surface soil moisture (θS) versus RI/RT and RS/RT in all subbasins. Increases in RI/RT were related to increases in θS, while increases in RS/RT occurred along with decreases in θS. These relations suggest that at the annual and seasonal scale, higher water availability in basins with dominant saturation-excess runoff result in increments of ET that deplete surface soil moisture. Overall, these established links are the outcome of interactions between soil, terrain, and rainfall properties that modulate runoff production and soil moisture for evapotranspiration.
d. Annual and monthly runoff ratios with comparison to empirical estimates
As part of our final analysis, we evaluated the runoff ratios (R/P) at annual, seasonal, and monthly scales to understand the implications of runoff mechanisms and evapotranspiration on streamflow. Figure 13 shows a comparison of runoff mechanism percentages (RI/RT, RS/RT, and RG/RT) during the summer (JAS) for subbasins BCN01 (dominated by RI) and BAV79 (with significant RG) in GAUGES and NLDAS ADJ scenarios. The comparisons are limited to summer conditions when large precipitation events occur. These subbasins were selected because they represent well the range of monthly runoff evolutions across the USR basin. As the NAM progresses from July to September, BCN01 decreases in RI/RT and increases in RS/RT for both scenarios, whereas the opposite trends are observed in BAV79. Moreover, RG/RT contributions in BAV79 increase as the monsoon progresses. This result implies that changes in antecedent wetness (θR in Fig. 10), as well as in rainfall intensity, differentially impact the runoff mechanisms in individual subbasins. As a result, we would expect the response in the entire USR basin to be an outcome of individual processes occurring in subbasins with a wide range of landscape properties and meteorological forcing. Table 4 presents monthly runoff mechanism percentages and associated runoff ratio fractions (RI/P, RS/P, and RG/P) for the USR basin, arranged from June 2007 to May 2008 for the different scenarios. Note the higher basin area contributing to runoff (AR) and the greater proportion of RI/RT for the NAM (JAS) relative to other months. In addition, the monthly infiltration-excess runoff ratio increases from RI/P = 0.01–0.02 (June) to RI/P = 0.03–0.16 (September) for the scenarios. An increase in monthly R/P during the NAM is consistent with Gochis et al. (2006) and suggests that infiltration-excess runoff is the major contributor to this process. In contrast, RS/P and RG/P play more important roles in the fall and winter seasons.
Percent contribution of runoff mechanisms (RI/RT, RS/RT, and RG/RT) during each summer month (JAS) for GAUGES and NLDAS ADJ in (a) BCN01 and (b) BAV79.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Monthly runoff mechanism percentages (RI/RT, RS/RT, and RG/RT) and monthly runoff ratio fractions (RI/P, RS/P, and RG/P) for the GAUGES, NLDAS RAW, and NLDAS ADJ scenarios. Here AR is the percentage of the USR basin that is contributing runoff during each month. The monthly runoff ratios are shown as the second row of each month, in italics. Dashed (–) symbols indicate when monthly precipitation was zero and a runoff ratio could not be calculated.
The spatial variability in annual and seasonal (DJF and JAS) runoff ratios is presented in Fig. 14 for the GAUGES and NLDAS ADJ scenarios. As a reference, we also show the annual runoff ratio (R/P) map in Fig. 14c used by the Mexican government for regional hydrologic studies (INEGI 2000b). The INEGI runoff coefficient estimates are based empirically on the soil permeability, vegetation cover, and the local precipitation intensity, as described in INEGI (2000b). A visual comparison of the annual runoff ratio maps suggests that NLDAS ADJ matches more closely the INEGI (2000b) estimates, which are defined based on a combination of soil and vegetation units (Figs. 1e,d). In particular, areas of low runoff ratio (R/P < 5%) coincide in the northern, eastern, and southern regions. GAUGES, on the other hand, generally overestimates the annual R/P, with certain eastern and western areas having R/P > 20%. As expected, the JAS distributions resemble the annual R/P map since most runoff events occurred in the NAM. In addition, most subbasins during JAS exhibit infiltration-excess runoff (RI/RT > 75%; hatched areas) as the dominant mechanism. While RI/RT is still a significant contributor to the DJF runoff ratios, certain subbasins with high RS and RG (Fig. 6) tend to be responsible for larger winter R/P. In GAUGES and NLDAS ADJ, subbasins with R/P > 20% are frequent in winter and certain eastern and western subbasins exhibit R/P > 50% during DJF. These magnitudes are due to baseflow production in subbasins with shallow water tables during the winter, despite the low rainfall amounts. In addition, a large winter rainfall event occurring within certain subbasins with high antecedent wetness carried over from the NAM is also responsible (see Fig. 10).
Spatial variability of annual and seasonal runoff ratios (R/P) for (a) GAUGES and (b) NLDAS ADJ scenarios. (top) Annual, (middle) winter (DJF), and (bottom) summer (JAS) R/P. Subbasins with annual or seasonal RI/RT > 75% are hatched in all maps. (c) Annual runoff estimates from INEGI (2000b) are shown for comparison purposes.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
Based on the spatial variations of the annual runoff ratio (R/P), we constructed frequency distributions for the scenarios and compared these to estimates from INEGI (2000b) in Fig. 15. Given the paucity of hydrologic data, INEGI (2000b) represents the best available (and official) runoff estimates valid at the same landscape scale as our simulations. As shown in Fig. 14c, the spatial distribution of INEGI (2000b) corresponds to an intersection of soil and vegetation units with varying runoff potentials. We aggregated these to individual subbasins to compare the simulations to the estimated distribution. While more detailed runoff patterns can be produced by the model (Ivanov et al. 2004), the subbasin representations were considered sufficient for the USR basin. Clearly, GAUGES overestimates the annual R/P because of the misrepresentations in using Thiessen polygons of a set of sparse rain gauges. NLDAS RAW captures well the low R/P in INEGI (2000b) but underestimates values of R/P > 5%. On the other hand, NLDAS ADJ produces a reasonable match with INEGI (2000b) in particular for R/P > 5%, because of the more realistic rainfall patterns adjusted with ground data. This suggests that adjusted NLDAS products have value as forcing to hydrologic models in remote regions with limited ground observations.
Frequency distribution of the annual runoff ratio, R/P (%), for GAUGES, NLDAS ADJ, and INEGI (2000b) in the USR basin.
Citation: Journal of Hydrometeorology 13, 3; 10.1175/JHM-D-11-0112.1
4. Synthesis and conclusions
This study documents an advanced technique for generating hydrologic predictions in mountainous basins under the influence of the NAM, characterized by variable atmospheric conditions. We evaluated the use of meteorological forcing from ground stations and from NLDAS and its adjustment. The scenarios yielded bimodal hydrologic forecasts, with 52%–78% of the annual runoff occurring during the NAM and a secondary winter runoff peak, which is consistent with Gochis et al. (2006) and Brito-Castillo et al. (2003). In addition, the simulations indicated that infiltration-excess runoff was the dominant mechanism, since low conductivity soils cover more than 50% of the area, though both saturation-excess runoff and groundwater exfiltration play important roles in specific regions. These outcomes are in agreement with the observed mixture of runoff mechanisms identified by Descroix et al. (2007). Furthermore, an analysis of the annual runoff ratios from the scenarios resulted in 14% (GAUGES), 3% (NLDAS RAW), and 6% (NLDAS ADJ), which compares well with estimates of 9%–19% from Gochis et al. (2006) and Viramontes and Descroix (2003). Similarly, the runoff ratios during the NAM (14%, 2%, and 4% for GAUGES, NLDAS RAW, and NLDAS ADJ) were consistent with values cited by Gochis et al. (2006) for wetter basins farther south in the region (9%–43%; average of 26%).
The distributed scenarios and hydrologic model afforded the opportunity of quantifying the spatiotemporal features of the watershed response in a large river basin over an annual cycle. Because of limited observations, the characteristics of the basin responses are poorly understood in the NAM region at these spatiotemporal scales (see Vivoni et al. 2009, 2010 for other attempts). We identified how precipitation patterns translated nonlinearly into water balance components, streamflow generation, and runoff mechanisms for a set of subbasins and the entire USR basin. Spatial expressions of different runoff mechanisms were directly related to terrain, soil, and rainfall properties. Analyses in individual subbasins revealed that the prevalence of infiltration-excess runoff depended primarily on soil characteristics, whereas saturation-excess runoff and groundwater exfiltration were influenced by terrain features, which is consistent with field studies of Descroix et al. (2002a,b). NLDAS ADJ provided the most appealing spatial variations in precipitation and resulted in an annual runoff ratio that best matched an empirical estimate by INEGI (2000b) in terms of frequency and spatial distributions. As a result, the procedure for adjusting the NLDAS fields with ground data yielded large improvements in their utility for hydrologic applications, as suggested by Luo et al. (2003). An outcome of coupling the adjusted NLDAS forcing and distributed model is the production of a best available hydrologic prediction system that starts to meet the needs of water resources planners in the NAM region in terms of agreement with official estimates and in providing details at fine spatiotemporal scales.
The main goal of this study was to explore the relationship between evapotranspiration and the underlying runoff mechanisms. Through the distributed simulations in the large, heterogeneous watershed, we found a previously undocumented link between the mode of runoff generation and the evapotranspiration fluxes that mediate boundary layer processes. We identified that the fundamental relation between soil moisture and evapotranspiration exhibited large differences among subbasins characterized by different runoff mechanisms for the NAM period. Subbasins with large saturation-excess runoff and groundwater exfiltration contributions exhibit more ET because of the availability of deeper root zone soil moisture. The forms of the simulated ET–θS relations were also consistent with an observed relation obtained at an eddy covariance tower in the Sonora River (Vivoni et al. 2010). Furthermore, we identified that the production of evapotranspiration at annual and seasonal scales was related to the proportion of individual runoff components. An increase in ET occurred when the fraction of saturation-excess runoff increased in a subbasin, whereas an opposite trend was found for the proportion of infiltration-excess runoff. The simulations conducted in this study were conducted for a normal precipitation year. We expect that for wetter years, higher fractions of saturation-excess runoff and groundwater exfiltration would be observed, leading to ET rates that would be sustained for longer periods. During drier years, a higher proportion of infiltration-excess with minimal contributions of saturation-excess runoff and groundwater exfiltration are expected. These relations suggest that difficult-to-observe spatial differences in runoff mechanisms could potentially be inferred from more readily available estimates of the evapotranspiration field (e.g., Mu et al. 2007) or the ecosystem distributions that directly influence the atmospheric exchange (e.g., Tang et al. 2012). The link between runoff mechanisms and evapotranspiration also suggests that the coupling between water and energy processes in the heterogeneous land–atmosphere interface is more complex than previously thought in seasonally wet hydrologic systems.
Acknowledgments
This work was supported by the Mexican Council for Science and Technology through a fellowship to A.R-M. We also thank NOAA Climate Program Office (Grant GC07-019) and NSF IRES Program (Grant OISE 0553852) for support provided to conduct field activities. The following individuals contributed to the model setup, simulations, and data processing: Katelyn Watson, Rudiger Escobar, Giuseppe Mascaro, Luis A. Méndez-Barroso, Emily Van Dam, Cara Shonsey, and Richard Vendlinski. We also appreciate the datasets provided by CEA and CONAGUA.
REFERENCES
Acutis, M., and Donatelli M. , 2003: SOILPAR 2.00: Software to estimate soil hydrological parameters and functions. Eur. J. Agron., 18, 373–377.
Arora, V. K., 2002: The use of the aridity index to assess climate change effect on annual runoff. J. Hydrol., 265, 164–177.
Batjes, N. H., 2002: Soil parameter estimates for the soil types of the world for use in global and regional modeling. International Food Policy Research Institute and International Soil Reference and Information Center Rep. 2002/02c, 52 pp.
Bedient, P. B., Huber W. C. , and Vieux B. E. , 2008: Hydrology and Floodplain Analysis. Prentice Hall, 816 pp.
Beven, K. J., and Kirkby M. J. , 1979: A physically-based variable contributing area model of basin hydrology. Hydrol. Sci. Bull., 24, 43–69.
Brito-Castillo, L., Douglas A. V. , Leyva-Contreras A. , and Lluch-Belda D. , 2003: The effect of large-scale circulation on precipitation and streamflow in the Gulf of California continental watershed. Int. J. Climatol., 23, 751–768.
Brown, D. E., 1994: Biotic Communities of the Southwestern United States and Northwestern Mexico. The University of Utah Press, 342 pp.
CEA, 2005: Estudio geo-hidrológico de las subcuencas de las Ríos Sonora, Zanjón, San Miguel, Mesa del Seri-La Victoria y cuenca Bacoachi. Comision Estatal del Agua Rep. CEA-ED-081 and CEA-ED-081-CA1, 391 pp.
Coblentz, D. D., and Riitters K. H. , 2004: Topographic controls on the regional-scale biodiversity of the southwestern USA. J. Biogeogr., 31, 1125–1138.
Cosgrove, B. A., and Coauthors, 2003: Real-time and retrospective forcing in the North American Land Data Assimilation System (NLDAS) project. J. Geophys. Res., 108, 8842, doi:10.1029/2002JD003118.
Daly, C., Neilson R. P. , and Phillips D. L. , 1994: A statistical-topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140–158.
Descroix, L., Gonzalez-Barrios J. L. , Vandervaere D. , Viramontes D. , and Bollery A. , 2002a: An experimental analysis of hydrodynamic behaviour on soils and hillslopes in a subtropical mountain environment (Western Sierra Madre, Mexico). J. Hydrol., 266, 1–14.
Descroix, L., Nouvelot J.-F. , and Vauclin M. , 2002b: Evaluation of an antecedent precipitation index to model runoff yield in the western Sierra Madre (North-west Mexico). J. Hydrol., 263, 114–130.
Descroix, L., Viramontes D. , Estrada J. , Gonzalez Barrios J. L. , and Asseline J. , 2007: Investigating the spatial and temporal boundaries of Hortonian and Hewlettian runoff in Northern Mexico. J. Hydrol., 346, 144–158.
Dominguez, F., Kumar P. , and Vivoni E. R. , 2008: Precipitation recycling variability and ecoclimatological stability—A study using NARR data. Part II: North American monsoon region. J. Climate, 21, 5187–5203.
Douglas, M. W., Maddox R. A. , Howard K. , and Reyes S. , 1993: The Mexican monsoon. J. Climate, 6, 1665–1677.
Eltahir, E. A. B., 1998: A soil moisture–rainfall feedback mechanism: 1. Theory and observations. Water Resour. Res., 34, 765–776.
FAO, 1988: FAO-UNESCO soil map of the world: Revised legend. World Soil Resources Tech. Paper 20, 140 pp.
Forzieri, G., Castelli F. , and Vivoni E. R. , 2011: Vegetation dynamics within the North American monsoon region. J. Climate, 24, 1763–1783.
García-Oliva, F., Maass J. M. , and Galicia L. , 1995: Rainstorm analysis and rainfall erosivity of a seasonal tropical region with a strong cyclonic influence on the Pacific coast of Mexico. J. Appl. Meteor., 34, 2491–2498.
Gebremichael, M., Vivoni E. R. , Watts C. J. , and Rodriguez J. C. , 2007: Submesoscale spatiotemporal variability of North American monsoon rainfall over complex terrain. J. Climate, 20, 1751–1773.
Gochis, D. J., Brito-Castillo L. , and Shuttleworth W. J. , 2006: Hydroclimatology of the North American Monsoon region in northwest Mexico. J. Hydrol., 316, 53–70.
Gochis, D. J., Watts C. J. , Garatuza-Payan J. , and Cesar-Rodriguez J. , 2007: Spatial and temporal patterns of precipitation intensity as observed by the NAME Event Rain gauge Network from 2002 to 2004. J. Climate, 20, 1734–1750.
Goodrich, D. C., Lane L. J. , Shillito R. M. , Miller S. N. , Syed K. H. , and Woolhiser D. A. , 1997: Linearity of basin response as a function of scale in a semiarid watershed. Water Resour. Res., 33, 2951–2965.
Gutiérrez, J., and Hernandez I. I. , 1996: Runoff and interill erosion as affected by grass cover in a semi-arid rangeland of northern Mexico. J. Arid Environ., 34, 287–295.
Hallack-Alegria, M., and Watkins D. W. , 2007: Annual and warm season drought intensity–duration–frequency analysis for Sonora, Mexico. J. Climate, 20, 1897–1909.
Hamlet, A. F., Mote P. W. , Clark M. P. , and Lettenmaier D. P. , 2007: Twentieth-century trends in runoff, evapotranspiration, and soil moisture in the western United States. J. Climate, 20, 1468–1486.
Higgins, R. W., Shi W. , Yarosh E. , and Joyce R. , 2000: Improved United States precipitation quality control system and analysis. NCEP/Climate Prediction Center ATLAS 7, 40 pp.
INEGI, 2000a: Síntesis de información geográfica del Estado de Sonora. Instituto Nacional de Estadisticas y Geografia Rep., 88 pp.
INEGI, 2000b: Cartas hidrologicas de aguas superficiales. Instituto Nacional de Estadisticas y Geografia Rep., 1 pp.
INEGI–INIFAP, 2001: Inventario nacional de suelos. Mexico Instituto Nacional de Estadisticas y Geografia and Instituto Nacional de Investigaciones Forestales, Agricolas, y Pecuarias Rep., 1 pp.
Ivanov, V. Y., Vivoni E. R. , Bras R. L. , and Entekhabi D. , 2004: Catchment hydrologic response with a fully distributed triangulated irregular network model. Water Resour. Res., 40, W11102, doi:10.1029/2004WR003218.
Joyce, R. J., Janowiak J. E. , Arkin P. A. , and Xie P. , 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487–503.
Koster, R. D., and Suarez M. J. , 1999: A simple framework for examining the interannual variability of land surface moisture fluxes. J. Climate, 12, 1911–1917.
Luo, L., and Coauthors, 2003: Validation of the North American Land Data Assimilation System (NLDAS) retrospective forcing over the southern Great Plains. J. Geophys. Res., 108, 8843, doi:10.1029/2002JD003246.
Luo, Y., Berbery E. H. , Mitchel K. E. , and Betts A. K. , 2007: Relationships between land surface and near-surface atmospheric variables in the NCEP North American Regional Reanalysis. J. Hydrometeor., 8, 1184–1183.
Mascaro, G., and Vivoni E. R. , 2010: Statistical and scaling properties of remotely sensed soil moisture in two contrasting domains in the North American monsoon region. J. Arid Environ., 74, 572–578.
Méndez-Barroso, L. A., and Vivoni E. R. , 2010: Observed shifts in land surface conditions during the North American Monsoon: Implications for a vegetation–rainfall feedback mechanism. J. Arid Environ., 74, 549–555.
Mitchell, K. E., and Coauthors, 2004: The multi-institution North American Land Data Assimilation System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system. J. Geophys. Res., 109, D07S90, doi:10.1029/2003JD003823.
Mora, F., and Iverson L. R. , 1998: On the sources of vegetation activity variation, and their relation with water balance in Mexico. Int. J. Remote Sens., 19, 1843–1871.
Mu, Q., Heinsch F. A. , Zhao M. , and Running S. M. , 2007: Development of a global evapotranspiration algorithm based on MODIS and global meteorology data. Remote Sens. Environ., 111, 519–536.
Muñoz-Arriola, F., Avissar R. , Zhu C. , and Lettenmaier D. P. , 2009: Sensitivity of the water resources of Rio Yaqui Basin, Mexico, to agriculture extensification under multiscale climate conditions. Water Resour. Res., 45, W00A20, doi:10.1029/2007WR006783.
Nesbitt, S. W., Gochis D. J. , and Lang T. J. , 2008: The diurnal cycle of clouds and precipitation along the Sierra Madre Occidental observed during NAME-2004: Implications for warm season precipitation estimation in complex terrain. J. Hydrometeor., 9, 728–743.
Paredes-Aguilar, R., Van Devender T. , and Felger R. S. , 2000: Cactaceas de Sonora, Mexico: Su Diversidad, Uso y Conservacion. IMADES Arizona-Sonora Desert Museum Press, 143 pp.
Pinker, R. T., and Coauthors, 2003: Surface radiation budgets in support of the GEWEX Continental-Scale International Project (GCIP) and the GEWEX Americas Prediction Project (GAPP), including the North American Land Data Assimilation System (NLDAS) project. J. Geophys. Res., 108, 8844, doi:10.1029/2002JD003301.
Robles-Morua, A., Mayer A. S. , Auer M. T. , and Vivoni E. R. , 2012: Modeling riverine pathogen fate and transport in Mexican rural communities and associated public health implications. J. Environ. Manage., in press.
Rodríguez-Iturbe, I., and Porporato A. , 2004: Ecohydrology of Water-Controlled Ecosystems. Cambridge University Press, 442 pp.
Sankarasubramanian, A., and Vogel R. M. , 2002: Annual hydroclimatology of the United States. Water Resour. Res., 38, 1083, doi:10.1029/2001WR000619.
Schapp, M. G., Leij F. J. , and van Genuchten M. T. , 2001: Rosetta: A computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. J. Hydrol., 251, 163–176.
Serrat-Capdevila, A., Valdes J. B. , Perez J. G. , Baird K. , Mata L. J. , and Maddock T. , 2007: Modeling climate change impacts—and uncertainty—on the hydrology of a riparian system: The San Pedro Basin (Arizona/Sonora). J. Hydrol., 347, 48–66.
Small, E. E., 2001: The influence of soil moisture anomalies on variability of the North American Monsoon System. Geophys. Res. Lett., 28, 139–142.
Smith, M. B., Seo D.-J. , Koren V. I. , Reed S. M. , Zhang Z. , Duan Q. , Moreda F. , and Cong S. , 2004: The Distributed Model Intercomparison Project (DMIP): Motivation and experiment design. J. Hydrology, 298, 4–26.
Steiner, M., Smith J. A. , Burges S. J. , Alonso C. V. , and Darden R. W. , 1999: Effects of bias adjustment and rain gauge data quality control on radar rainfall. Water Resour. Res., 35, 2487–2503.
Tang, Q., Vivoni E. R. , Muñoz-Arriola F. , and Lettenmaier D. P. , 2012: Predictability of evapotranspiration patterns using remotely sensed vegetation dynamics during the North American monsoon. J. Hydrometeor., 13, 103–121.
Viramontes, D., and Descroix L. , 2003: Changes in the surface water hydrologic characteristics of an endoreic basin of northern Mexico from 1970 to 1998. Hydrol. Processes, 17, 1291–1306.
Vivoni, E. R., Ivanov V. Y. , Bras R. L. , and Entekhabi D. , 2004: Generation of triangulated irregular networks based on hydrological similarity. J. Hydrol. Eng., 9, 288–302.
Vivoni, E. R., Ivanov V. Y. , Bras R. L. , and Entekhabi D. , 2005: On the effects of triangulated terrain resolution on distributed hydrologic models. Hydrol. Processes, 19, 2101–2122.
Vivoni, E. R., Entekhabi D. , Bras R. L. , and Ivanov V. Y. , 2007a: Control on runoff generation and scale-dependence in a distributed hydrologic model. Hydrol. Earth Syst. Sci., 11, 1683–1701.
Vivoni, E. R., Gutiérrez-Jurado H. A. , Aragon C. A. , Méndez-Barroso L. A. , Rinehart A. J. , and Wyckoff R. L. , 2007b: Variation of hydrometeorological conditions along a topographic transect in northwestern Mexico during the North American monsoon. J. Climate, 20, 1792–1809.
Vivoni, E. R., Moreno H. A. , Mascaro G. , Rodríguez J. C. , Watts C. J. , Garatuza-Payan J. , and Scott R. L. , 2008: Observed relation between evapotranspiration and soil moisture in the North American monsoon region. Geophys. Res. Lett., 35, L22403, doi:10.1029/2008GL036001.
Vivoni, E. R., Tai K. , and Gohis D. J. , 2009: Effects of the initial soil moisture on rainfall generation and subsequent hydrologic response during the North American monsoon. J. Hydrometeor., 10, 644–664.
Vivoni, E. R., Rodríguez J. C. , and Watts C. J. , 2010: On the spatiotemporal variability of soil moisture and evapotranspiration in a mountainous basin within the North American monsoon region. Water Resour. Res., 46, W02509, doi:10.1029/2009WR008240.
Vivoni, E. R., Mascaro G. , Mniszewski S. , Fasel P. , Springer E. P. , Ivanov V. Y. , and Bras R. L. , 2011: Real-world hydrologic assessment of a fully-distributed hydrological model in a parallel computing environment. J. Hydrol., 409, 483–496.
Wood, E. F., and Coauthors, 2011: Hyperresolution global land surface modeling: Meeting a grand challenge for monitoring Earth’s terrestrial water. Water Resour. Res., 47, W05301, doi:10.1029/2010WR010090.
Xu, J. J., Shuttleworth W. J. , Gao X. , Sorooshian S. , and Small E. E. , 2004: Soil moisture–precipitation feedback on the North American monsoon system in the MM5-OSU model. Quart. J. Roy. Meteor. Soc., 130, 2873–2890.
