Sensitivity of the water resources of Rio Yaqui Basin, Mexico, to agriculture extensification under multiscale climate conditions



[1] The Yaqui River Basin (YRB) is the most important wheat-producing region in Mexico. The main source of irrigation water in the Yaqui basin (over 80%) is surface water. Crop production sustainability is therefore closely linked to YRB streamflow and potentially to its sensitivity to climate variability and land use changes. We study the sensitivity of streamflow to land cover change resulting primarily from conversion of natural vegetation to unirrigated agriculture within the basin. We also examine how this sensitivity is influenced by midscale (North American Monsoon) and large-scale (El Niño–Southern Oscillation (ENSO) and Pacific Decadal Oscillation (PDO)) climate regimes. Streamflow simulations using the variable Infiltration Capacity Model were performed in which current land use was altered incrementally within the three major subbasins of the YRB. In general, we found that the sensitivity of basin runoff to ENSO-and PDO-related variations in precipitation was much greater than to land cover–related changes and was highest during winter. Furthermore, climate and land cover runoff sensitivities appear to be related; in El Niño (generally wet) years the sensitivity of streamflow to land cover change can be four times higher than in La Niña (generally dry) years. The sensitivity of streamflow to PDO phase was smaller. Streamflow sensitivities to climate were the highest when PDO and ENSO were in phase. We conclude that climate variability exerts stronger controls on the surface hydrology than does land use change associated with the extensification of agriculture. For instance, flows in El Niño versus La Niña years differed by 76% in winter and 16% in summer, whereas maximum monthly (winter) flow changes were at most 4%.

1. Introduction

[2] Land use change has been increasingly recognized over the last thirty years as a significant driver of changes in the land surface water and energy balances. Projects such as the land use and land cover change (LUCC) program element of the International Geosphere-Biosphere Program (IGBP), have studied the dynamics of land use changes as well as their impact on climate at multiple scales [Lambin et al., 2006; Mahmood et al., 2006]. The expansion of cropland produced by clearing the natural land cover is known as agricultural extensification and has been a major form of global land use change at different spatial scales and during different stages of human cultural evolution [Tilman, 1999; Lambin et al., 2003]. For instance, Goldewijk and Ramankutty [2004] report a 50% net increase in agricultural areas globally in the 20th century. In consequence, the hydrosphere's sensitivity to land use changes is closely linked to agricultural extensification and deforestation, which impact the land-atmosphere-ocean water and energy fluxes at multiple spatial scales. At the regional scale, land use changes can modify atmospheric mesoscale circulations and redistribute precipitation [Avissar and Liu, 1996; Weaver and Avissar, 2001; Baidya Roy and Avissar, 2002; Ramos da Silva and Avissar, 2006; Chhabra and Geist, 2006]; while at the global scale, the impacts can alter climate through teleconnections [Werth and Avissar, 2005]. With respect to the land surface hydrological cycle, land use changes alter thermal and dynamic properties of the atmospheric boundary layer through the enhancement or inhibition of evapotranspiration and soil moisture dynamics [Milly et al., 2005, D'Odorico and Porporato, 2006]. The impacts of land use change on surface hydrology produced by agricultural extensification and accompanying deforestation have been examined through a combination of hydrological modeling and retrospective data analysis [Lorup et al., 1998; Braud et al., 2001; Weber et al., 2001; De Roo et al., 2001; Fohrer et al., 2001; Klocking and Haberlandt, 2002; Wegehenkel, 2002; Kokkonen and Jakeman, 2002; Bronstert and Bardossy, 2003; Bronstert, 2004; Twine et al., 2004; Legesse et al., 2003; Cosandey et al., 2005; Fennell et al., 2006; Claessens et al., 2006; Zhang and Schilling, 2006].

[3] In surface hydrology, the response of a watershed's runoff to precipitation is related both to infiltration, which controls the “fast response” of a watershed immediately following precipitation events, and to deep soil moisture and/or groundwater, which controls the “slow response.” Infiltration depends on near-surface soil moisture, and hence on antecedent evapotranspiration (ET) prior to a storm event. Runoff is also generated between storms by slow runoff response, or base flow, which depends on the accumulated infiltration, as well as antecedent ET from deep soils (e.g., by trees). Long-term observational studies have shown increased streamflow following deforestation, mostly as a result of reduced evapotranspiration, but also because of changes in infiltration resulting from altered soil hydraulic properties [De Roo et al., 2001; Klocking and Haberlandt, 2002].

[4] In southwestern North America (SWN) (see Figure 1) land use changes in recent decades have resulted from increasing population (southwestern United States) and agricultural development (northwestern Mexico). Our hypothesis is that the impact of agricultural extensification on the water resources of SWN may be modulated by climate variability. The North American Monsoon (NAM), which occurs between about June and September, is responsible for 50–80% of the annual precipitation in SWN. Similarly, midscale and large-scale climate phenomena, specifically El Niño–Southern Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO), affect the precipitation regimes in the region, producing wetter or drier monsoon seasons [Castro et al., 2001; Higgins and Shi, 2001; Englehart and Douglas, 2004, 2006]. For example, Lau et al. [2004] and Englehart and Douglas [2006] mentioned that positive PDO is associated with above normal precipitation during the monsoon season, while Mantua et al. [1997] generalize that PDO is positively correlated with winter precipitation over the northern Mexico. The Yaqui River Basin (YRB) lies within the domain of influence of NAM (see Figure 1). Its surface hydrology is strongly affected by seasonal and interannual precipitation variability, especially as related to NAM. These patterns define the spatial distribution of runoff within the basin, which in turn controls YRB streamflow, some 90% of which is consumed for irrigation and other human use. Hence, changes in the streamflow of the Yaqui River directly affect crop yield from the irrigated portion of the basin, most of which is in the lower portion of the basin downstream of the Oviachic Dam (see Figure 1). The extent to which changes upstream of Oviachic could affect the sustainability of major agricultural developments located downstream, in the Yaqui Valley, is unclear.

Figure 1.

Yaqui River Basin, including channel network, major subbasins, and stream gauge locations.

[5] It should be noted that while the irrigated portion of the basin constitutes only about 3% of its total drainage area, this irrigated region is critical to the socioeconomic health not only of the region, but of the entire country [Naylor et al., 2001; Luers et al., 2003]. Nonetheless, land cover change has occurred over a much larger portion of YRB, much of which is unirrigated, and it is the sensitivity of YRB streamflow to land cover change in this much larger area that is investigated in this paper. For instance, Instituto Nacional de Estadística Geografía e Informática (INEGI) ( = geo) suggests that potential changes in current vegetation due to the extensive agricultural extensification may affect around 10% of the Yaqui basin, and much of this area is located upstream of the Oviachic Dam.

[6] The interaction of climate variability over a range of temporal scales from interannual (NAM, essentially interannual variability in the onset and strength of the monsoon), and multiyear to decadal (ENSO, multiple years and PDO, multidecadal) with the effects of vegetation change are not well known. The objective of this study is to evaluate the sensitivity of streamflow to land use changes in the YRB under different climatological regimes. To accomplish this objective, we apply the variable infiltration capacity (VIC) macroscale hydrology model, which accounts for the exchanges of water and energy using an explicit representation of the subgrid distribution of vegetation. The structure of VIC is well suited for examination of the effects of vegetation change on river basin hydrologic response, and it has been previously applied for this purpose in the Columbia River basin [Matheussen et al., 2000; Haddeland et al., 2002; VanShaar et al., 2002] as well as over most of the northern hemisphere [Haddeland et al., 2007]. We ran VIC over YRB for the 50 year period 1949–1999, a period of sufficient length to allow examination of the influence of NAM, ENSO, and PDO variability on the basin's hydrologic response as related to long-term vegetation change.

2. Methodology

[7] The VIC model [Liang et al., 1994, 1996] contains an explicit representation of the effects of vegetation on land-atmosphere energy and moisture fluxes. We used the model to evaluate possible impacts of gradual increases in (unirrigated) crop areas by running a series of historical simulations. We used a 50 year (1949–1999) gridded forcing data set for the NAM region at 1/8° latitude-longitude spatial resolution produced by Zhu and Lettenmaier [2007]. It includes 3-hourly precipitation, surface wind, surface air temperature, and downward and longwave radiation, all of which are required to force the VIC model. The gridded data are based on daily observations of precipitation interpolated to 3-hourly intervals from the station archive of Servicio Meteorologico Nacional. The data were gridded using methods described by Maurer et al. [2002]. It is worth noting that the 3-hourly precipitation was disaggregated from daily data by assuming a constant average precipitation rate throughout the day. This might be expected to pose problems for simulations in northwestern Mexico, where the diurnal cycle of precipitation plays an important role in the surface hydrology [Gochis et al., 2004]. However, as shown by Maurer et al. [2002] the runoff predicted by VIC at regional and monthly scales shows only weak sensitivity to different methods of disaggregating daily precipitation (the diurnal cycle of solar radiation is far more important, and is the main reason for running the model at a subdaily time step). Wind speed data were taken from the NCEP-NCAR reanalysis [Kalnay et al., 1996]. Surface air temperature data were gridded from observations; humidity and downward shortwave and longwave radiation were derived from the daily temperature range and daily mean temperature following methods summarized by Maurer et al. [2002]. Soil features were taken from the 5 min resolution Food and Agriculture Organization data set [Food and Agriculture (FAO), 1995]. The associated soil characteristics were obtained using the formulations of Cosby et al. [1984], Rawls et al. [1998], and Reynolds et al. [2000]. Vegetation parameters were taken from the University of Maryland global land cover data set [Hansen et al., 2000].

[8] The model was run in full energy balance mode as in the work by Maurer et al. [2002], meaning that the components of the hydrological cycle were produced every 3 h by iteratively solving the surface energy balance for an effective surface temperature. Model calibration was performed manually, focusing on the following parameters of the model's three-layer soil column: the soil moisture capacity curve shape parameter (bi), the maximum base flow rate of the third soil layer (dsm), the fraction of the maximum third-layer soil moisture at which base flow becomes nonlinear (ws), and the thickness of the second soil layer (s2). bi determines infiltration and the three other parameters define base flow (slow runoff) response and the water holding capacity of the two lower soil layers. The routing model of Lohmann et al. [1998a, 1998b] was used to route the runoff and base flow from individual grid cells through the basin's channel network and to simulate daily streamflow at three gauges within the basin: Angostura, Paso Nacori, and Oviachic (see Figure 1). Streamflow observations were obtained from Banco Nacional de Datos de Aguas Superficiales [2004]. Figure 2 shows VIC model performance relative to streamflow observations using the modified parameters for a 10 year calibration and 5 year validation periods. Aside from a positive bias for peak flows at Oviachic, the model plausibly reproduces observations at the three stations.

Figure 2.

Streamflow observations (dashed line) versus modeled performance (solid line) for the Angostura (0.69, 0.48), Paso Nacori (0.75, 0.77), and Oviachic (0.75, 0.83) basins in m3 s−1; Nash-Sutcliffe efficiency index for the calibration and the validation is given in parentheses. The calibration process was based on monthly simulated and observed values for the calibration period of January 1968 to December 1978 and the validation period of January 1979 to December 1984.

3. Site Description

[9] The Yaqui River Basin (73,000 km2) is located between 34°N and 32°N in the Mexican states of Sonora and Chihuahua and part of Arizona in the United States (see Figure 1). From east to west, the prevalent slope varies from strong to weak as the river flows from its headwaters to the Gulf of California. The Yaqui River (main stem length 397 km) is formed by the confluence of the Bavispe (374 km) and Aros-Papigochi (125 km) rivers, both of which originate in the Sierra Madre Occidental. Downstream it merges with the Moctezuma River. Twenty hydrometric stations are located along the Yaqui River system. The Angostura, Paso Nacori, and Oviachic stations (see Figure 1) were selected for use in this study based on features such as at most modest effects of upstream water management, outflow location, and continuity and reliability of the streamflow data record. The Angostura station is located on the Rio Bavispe above the Lazaro Cardenas Dam. This station accounts for the natural flows drained from portions of southern Arizona, Sonora, and Chihuahua. Paso Nacori is located on the Aros Papagochic River above its confluence with the main stem Yaqui River. Flows at both Paso Nacori and Angostura are only modestly affected by upstream management. The Oviachic station is located above the Alvaro Obregon Dam. This station measures discharge from the entire YRB drainage area below two large reservoirs and about 20 smaller reservoirs. The discharge record for this station has been naturalized based on work by Emerson [2005] to account for upstream storage and diversions. As noted above, about 3% of the YRB drainage area, or 2250 km2, is irrigated, and most of this area is below the Oviachic station.

4. Climatic Regime Identification

[10] The effects of ENSO were evaluated through the use of the bivariate ENSO time series (BEST) index, which is based on the SST anomalies (El Niño 3.4) and sea level pressure differences between Tahiti and Darwin in the East and West Pacific coasts [Smith and Sardeshmukh, 2000]. The PDO index was defined as the first principal component of the North Pacific sea surface temperature on a monthly basis following Mantua et al. [1997]. PDO has a warm phase, which is characterized by negative SST anomalies in the northwestern Pacific and corresponding positive SST anomalies in the eastern tropical Pacific, and a cold phase with the opposite conditions. For consistency we use the BEST and PDO indices cited above for purposes of forming the composites described below.

[11] The main patterns of climate variability were defined as three ENSO phases: El Niño (EN), La Niña (LN), and neutral (N) years, each for warm and cold PDO phases. The effects of land use changes on streamflow were evaluated through formation of streamflow composites during these periods (Table 1).

Table 1. El Niño and La Niña Periods From 1949 to 1999 Based on the BEST Indexa
La NiñaEl Niño
  • a

    The gaps between El Niño and La Niña periods represent the neutral periods. The dates given in bold are cold PDO (1949–1977), and the other periods correspond to warm PDO (1978–1999).


[12] Streamflow composites were formed for monsoon and winter periods and for each combination of ENSO and PDO phases. The monsoon period was defined based on typical monsoon onset and end dates in Mexico portion of the basin (JJAS) [Douglas et al., 1993; Higgins et al., 2003]. Winter months were defined as DJFM.

[13] The streamflow composites were normalized by drainage area, for monsoon and winter months, in each climate category. EN and LN were based on the BEST index as defined by Smith and Sardeshmukh [2000]. Non-EN and -LN years were considered to be ENSO neutral (N) as shown in Table 1. Cold and warm PDO periods were taken to be 1949–1977 and 1978–1999, respectively, following Mantua et al. [1997]. Changes in streamflow resulting from climate variability were tested using the rank sum test [Berry, 1996] at 5% significance level (two-sided test).

5. Land Use Change Scenarios

[14] The land use change scenarios consisted of seven simulations of gradually increasing increments of (unirrigated) crop area. As noted above, our focus is on changes in vegetation and streamflow in the unirrigated areas upstream of Angostura, Paso Nacori, and Oviachic. Above Angostura, land cover historically has been dominated by grass and shrubland, and above Paso Nacori by forest (see Figure 3a). The basin upstream of Oviachic but downstream of Angostura and Paso Nacori has both land cover types, but in different proportions than above Angostura and Paso Nacori.

Figure 3.

(a) Current land use fraction and (b) distribution of the different vegetation types and distribution of the crop fraction (1, evergreen narrow leaf; 2, evergreen broadleaf; 3, deciduous narrow leaf; 4, deciduous broadleaf; 5, mixed forest; 6, woodland; 7, wooded grassland; 8, closed shrubland; 9, open shrubland; 10, grass; 11, crop; 12, bare soil; 13, urban surface).

[15] Figure 3a shows the predominant current vegetation types (as of 1992–1993 [Hansen et al., 2000]) in the YRB. We specified eight scenarios corresponding to original (current land use) and the replacement of the original vegetation by 1, 2, 3, 4, 5, 10, and 25% crop-type vegetation above Oviachic, using the INEGI ( (based on INEGI [2001]) data set as a baseline. Each grid cell in VIC was allowed to have a maximum of 12 different vegetation types plus bare soil. Starting with the original scenario, each vegetation type represented in a grid cell (except bare soil) was reduced by the percentages mentioned above, and these transitions were assumed to be to unirrigated cropland. Figure 3b shows the distribution of the main vegetation types and the distribution of crop fraction for each grid cell for current conditions and after increments of 10% and 25% of crop surfaces. For simplicity, we focus on the increment of 10% cropland. In Angostura, Paso Nacori, and Oviachic this increment corresponds to (new) crop fractions of 11%, 12%, and 13%, respectively. Figure 3b also shows that in some grid cells upstream of Oviachic crop fractions can be as high as 40% for our highest crop fraction scenario.

6. Results

6.1. Multiscale Climatology and Agricultural Extensification

[16] Agriculture extensification and deforestation in Mexico are closely related activities and define some of the major changes in land use over northwestern Mexico. FAO [2007] has determined annual deforestation rates over Mexico of 0.51% per year. These rates contrast with roughly equilibrium conditions in Canada and afforestation in the United States. Over northwestern Mexico, INEGI ( estimated that land use changes caused by agricultural extensification upstream of Oviachic have affected about 10% of the YRB. Such changes are mainly due to the development of unirrigated agricultural areas. Some studies have identified land use changes in basins of northwestern Mexico due to invasive species, agriculture, and urbanization, affecting the stability of watersheds [Kepner et al., 2000]. In the following sections we evaluate how these land use transitions might affect runoff for different climatic regimes. For simplicity we specify vegetation changes in terms of crop increments of 10%, which correspond roughly to the magnitudes of extensification of agricultural practices reported by INEGI ( We evaluate streamflow changes due to a range of increments in the crop surface fraction under different climate regimes (monsoon/winter, ENSO, and PDO).

6.1.1. Genesis and Evolution of NAM

[17] Douglas et al. [1993], Adams and Comrie [1997], and Higgins et al. [1997, 2003] have identified the genesis and evolution of the NAM by explaining the spatial variability of monsoon onset and the fraction of annual precipitation that falls in summer. Both characteristics vary among the subbasins of the YRB. For example, Figure 4a shows that precipitation in Angostura is 1–2 mm/d lower than in Paso Nacori during the summer but during winter, precipitation values are similar in all subbasins. A similar seasonal dependence is visible in the variability of streamflow across subbasins; Angostura summer (normalized) flows are substantially lower than those of Paso Nacori and Oviachic but winter differences among subbasins are less pronounced (see Figures 4b and 4c). These patterns also coincide with the subregions defined by Comrie and Glenn [1998], who identified four different spatial patterns of precipitation in the NAM region: two of their subregions encompass Paso Nacori and southern portion of Oviachic in the south and Angostura in the north, respectively.

Figure 4.

Climatological monthly values (mean for all climate categories) for (a) precipitation (in mm/month), (b) normalized streamflow (in mm/month), and (c) streamflow (in 106 m3/month). (d) Angostura normalized streamflow difference between current and altered conditions for crop increments of 1%, 2%, 3%, 4%, 5%, and 10% (mm/month). (e) Same as Figure 4d but for Paso Nacori. (f) Same as Figure 4d but for Oviachic.

[18] The model-predicted effect of agricultural extensification on streamflow resulting from prescribed changes to the current spatial distribution of vegetation as described in section 5 is shown in Figures 4d4f. Streamflow differences between current and modified land use conditions (for increments of 10% in crop surface) in Angostura, Paso Nacori, and Oviachic show two main patterns of variation throughout the year (Figures 4d4f). The first pattern of variability occurs during the monsoon months and consists of increases of streamflow up to 0.13 mm/month in response to increased crop area. The second pattern of streamflow variability is observed during winter, and consists of decreases in streamflow of as much as 0.4 mm/month in response to increased crop area. These patterns are also observed in Paso Nacori and Oviachic but with different durations and streamflow magnitudes. In Paso Nacori, the predicted signature of land cover change in monsoon months is stronger than in winter (Figure 4e). Increments in the crop areas result in streamflow increases of up to 0.64 mm/month in July. In winter, Paso Nacori streamflow reductions resulting from agriculture extensification are considerably smaller than those observed in Angostura. In fact, Paso Nacori's streamflow increases extend from the monsoon months up to March of the following year. Paso Nacori's streamflow reductions last only three months and are almost a factor of three less than those in Angostura. Simulated land cover change effects above Oviachic exhibit responses that are intermediate between those of Angostura and Paso Nacori (Figure 4f). Oviachic's summer streamflow increases by up to 0.4 mm/month between June and February, with a maximum in July. Oviachic's winter streamflow decreases by up to 0.2 mm/month, with a duration of four months and a maximum in March.

[19] The surface hydrological response in the YRB to increments in crop areas is different among the subbasins. For example, some of the changes in simulated streamflow highlight the controls exerted by vegetation and soil physics. In general, changes in streamflow generation are attributable to the large amplitude of the seasonal crop LAI variation in comparison to that of the natural vegetation that it replaces. In subbasins containing substantial forest or grass cover (e.g., Paso Nacori and Oviachic; see Figure 3), crop LAI may be comparable to that of the natural vegetation. Increasing crop area in these subbasins tends to decrease winter ET and increase winter streamflows as a result. In subbasins such as Angostura that contain less forest cover and more grass and shrub cover, crop LAI is comparable to or greater than that of the natural vegetation even in winter, and therefore increasing crop area tends to increase winter ET and decrease winter streamflows. Winter precipitation and soil properties may also contribute to the predicted increase in streamflow. This distinction between controls was described by Yildiz and Barros [2007] over a midlatitude basin, which was controlled alternatively by vegetation during summer and by soil physics during spring.

[20] Table 2 shows the weighted average of vegetation fraction and LAI (the effective spatial average LAI) during the monsoon and winter months for the subbasins of the YRB, under current and agriculturally extensified conditions. The change in this parameter, which we term the vegetation parameter (VP), can be used to understand the effects of agriculture extensification relative to current vegetation at the basin level. During monsoon months for all subbasins, the modified land use VP was higher than the current land use VP, which was reflected in reduced streamflow generation. In winter, for Angostura and Oviachic basins, modified land use VP increased and streamflow decreased, however in the Angostura basin, despite a small reduction in VP, there was a substantial increase in streamflow. The magnitude of the increase, which at first glance appears to be inconsistent with the magnitude of the vegetation change, has to do with spatial variations in soil properties and precipitation; the areas within the basin where there is the greatest increase in LAI are also the areas that are disproportionately responsible for runoff generation within the basin, and therefore the effect of changes in VP basinwide tends to be amplified.

Table 2. Vegetation Parameter VPa
 AngosturaPaso NacoriOviachic
  • a

    VP is defined as the product of the vegetation fraction and LAI for monsoon and winter months. Percentage of change between current and modified conditions streamflows are given in parentheses.

10%5.4 (0.33)7.1 (1.55)6.2 (1.17)
10%3.2 (−2.27)4.6 (1.06)3.8 (−0.31)

6.1.2. ENSO

[21] The interaction of ENSO and land cover change in the YRB provides some insights into the relative effects of large-scale phenomena as contrasted with local to regional land use changes. The multiannual variability of simulated streamflow over the YRB is generally consistent with precipitation variability elsewhere in the NAM region as reported by Higgins et al. [1999], Castro et al. [2001], and Gutzler [2004]. The largest streamflows in the Angostura, Paso Nacori, and Oviachic subbasins in EN conditions occur during winter while the largest streamflow values in LN conditions occur during summer as shown by the composites in Figure 5. In Angostura, the streamflow differences between EN and LN (for all years) were on average 80 mm during the winter and −9 mm during the summer; neutral year streamflows were always intermediate in value between the EN and LN streamflows. In contrast, streamflow differences between EN and LN above Paso Nacori were on average 25 mm during winter and close to zero during the monsoon. In this particular case, positive differences between EN and LN were observed during the cold PDO phase and negative differences during the warm PDO phase (both statistically significant), as discussed in detail in section 6.1.3. As in Angostura, in the Oviachic basin EN streamflows were less than for LN during the winter and the opposite is true during the monsoon with average differences of 33 mm and −10 mm, respectively. Monsoon season streamflow differences between EN and LN were both statistically significant during the warm phase of PDO in Angostura and Oviachic. This difference suggests a triggering effect of warm PDO on ENSO precipitation patterns as suggested by Englehart and Douglas [2002] and Gutzler [2004]. Gutzler [2004] identified a synergistic effect between the warm PDO and ENSO through the increase in the correlation between EN 3.4 SST anomalies and precipitation anomalies. ENSO events and monsoon precipitation are negatively correlated under warm PDO [Englehart and Douglas, 2002], also shown by marked and mild influences of ENSO on the streamflow during winter and monsoon seasons, respectively.

Figure 5.

Normalized streamflow composites (mm/month) for the period 1949–1999 in (top) Angostura, (middle) Paso Nacori, and (bottom) Oviachic subbasins. LN, La Niña; EN, El Niño; N, neutral; C, cold PDO; W, warm PDO.

[22] The sensitivity of streamflow to land use changes for monsoon and winter months as modulated by ENSO is shown in Figure 6. The results are for 10% increments in crop surfaces; positive differences indicate that increments in crop surfaces produce an increase in streamflow, while negative differences indicate streamflow decreases.

Figure 6.

(top) Composite mean streamflow and (bottom) streamflow changes between current vegetation and 10% more crop area in mm per season (winter is DJFM and monsoon is JJAS) for La Niña (LN), El Niño (EN), and ENSO neutral (N) under the influence of cold PDO (C) and warm PDO (W) for (a and d) Angostura, (b and e) Paso Nacori, and (c and f) Oviachic.

[23] Across all subbasins and all climate categories, streamflow changes due to increasing crop area are less positive in winter than in monsoon months due to the temporal variability of LAI (Figure 6). Streamflow changes in winter during EN are the least positive or even negative, across all basins (e.g., −1.51 mm in Angostura) due to the changes in LAI and combined with high precipitation. Streamflow generation may result from the influence of changing infiltration capacities, which result from increasing precipitation excess and are regulated by evapotranspiration and soil moisture storage. In Angostura, streamflow changes are substantially more negative than for the other two basins for all climate categories. This is also due to the predominance of grass cover, which in the extensification scenario is replaced by crops with higher LAI, which during winter produce more ET, as discussed in section 6.1.1. As a result in Angostura, the reduction of streamflow caused by agricultural extensification is most pronounced during EN, with a major impact during winter. Hence, the main impacts occur during the period of recharge, which provides water for downstream irrigation during the following crop season. Analogous responses were observed by Lorup et al. [1998] in semiarid catchments, where increments of urban and crop surfaces in areas originally dominated by woodland/grassland reduced runoff production. On the other hand, Paso Nacori's major changes occur during the monsoon (differences >0.56 mm). In general, streamflow differences between EN and LN were more pronounced during winter (>0.3 mm) than during the monsoon (<0.2 mm).

6.1.3. Long-Term Influences

[24] Streamflow differences between cold PDO and warm PDO phases are larger during winter than during the monsoon (Figure 6). In fact, Angostura streamflow differences between cold PDO and warm PDO tend to be larger during EN (38.7 and 71 mm for cold and warm PDO) than during LN (8.6 and 18.7 mm for cold and warm PDO). Because Angostura is located in the northernmost portion of the YRB, these results are consistent with the warm PDO-related high precipitation over northern Mexico during the winter (also reported by Mantua et al. [1997]). Over the southern subbasins, streamflow differences associated with PDO are smaller and in the case of Paso Nacori, higher winter streamflow values occur during the cold PDO under both EN and LN conditions. Brito-Castillo et al. [2003] also identified the influence of the PDO on the streamflows over the NAM, using a tree ring reconstructed streamflow data set and assessing streamflow records over the river basins draining to the Gulf of California (not including the YRB).

[25] In the YRB winter streamflow differences were always more negative under warm PDO for Angostura and Oviachic, while Paso Nacori maintained the same positive differences in both summer and winter. Generally, the PDO is recognized as a climate phenomenon that primarily influences winter climate, however, the PDO may influence some components of monsoon physics, such as surface temperature contrasts between land and sea, which could furthermore influence the strength of the monsoon as suggested by Englehart and Douglas [2002, 2006] and Gutzler [2004].

[26] In addition to the simulations reported above, we performed additional sensitivity tests to evaluate the effects of replacement of forested areas by crops (not shown here). The surface hydrological responses were of the same order of magnitude in the winter and monsoon seasons as those discussed above. In general, the amplification of surface hydrologic response to the warm PDO in all subbasins during summer showed more homogeneous and higher responses than for cold PDO. In winter, similar hydrologic responses occurred under the influence of LN during the same warm PDO. However, during the cold PDO and under the influence of EN, streamflow differences were higher than for warm PDO.

6.2. Water Balance

[27] Streamflow is an integrator of hydroclimatological forcings at the basin scale [Sivapalan et al., 2003]. The sensitivity of streamflow to agricultural extensification under the influence of NAM, ENSO, and PDO was evaluated as a function of the changes in the components of the water cycle in the YRB subbasins. The surface water balance is

equation image

where ΔS is the change in the water storage in a determined area during the period Δt, P is precipitation, R is (surface) runoff, E is evapotranspiration, and B is base flow.

[28] Abdulla et al. [1996], Metcalfe and Buttle [1999], Laio et al. [2001], and Porporato et al. [2004] have used the water balance approach to show the effects of environmental changes on the hydrological cycle. The analysis of the components of the water balance allows identification of where the effect of land use changes is more important and how it contributes to streamflow generation.

[29] In addition to differences in vegetation cover, differences in soil characteristics play an important (but generally lesser) role in the subbasin responses to climate variability. For example, soil moisture is affected by soil permeability, which influences the temporal variability of base flow originating in the deepest VIC soil layer (layer 3). Figure 7 shows that decrease in base flow as a result of increasing crop area are larger (0.008 mm) in Angostura than in the central part of the YRB (0.003 mm). This spatial distribution of changes follows the spatial distribution of values of the VIC maximum base flow parameter dsm among the different subbasins (Angostura > Oviachic > Paso Nacori) and implies differences in permeability in the soils of these subbasins. Accordingly, moisture in the bottom soil layer follows the reverse pattern (Paso Nacori > Oviachic > Angostura). According to Yildiz and Barros [2007] soil physics exert major control on surface hydrology during the dry season and on ET during the wet season. Also, D'Odorico and Porporato [2006] note that in arid domains, evapotranspiration can be affected to a higher degree by soil moisture than by the evaporative demand (i.e., these systems are moisture limited). In the YRB, the dominant (agricultural) vegetation that we substitute for the native vegetation can exert important controls on soil moisture. Angostura is characterized by permeable soils and a variety of vegetation types including grassland, shrubland, and woodland [INEGI, 1993]. In the central part of YRB (including Paso Nacori and Oviachic), soils are characterized by middle to low permeability, and shrubs and grasses dominate vegetation. Toward the northwest, the permeability of the soils increases and the vegetation is characterized by Sinaloan deciduous forest, with intrusions of tropical elements [INEGI, 1993; Brown, 1994]. Nonetheless, the ET differences between modified and current conditions show similar responses. The base flow and soil moisture temporal changes are higher in Angostura than any other subbasin during winter (Figure 7). This produces a pronounced reduction in runoff and base flow that is reflected in the negative streamflow changes (see Figures 4d4f).

Figure 7.

Water balance components under current conditions and changes in water balance (obtained from the differences between altered conditions (crop increment of 10%) and current conditions). O, Oviachic; P, Paso Nacori; A, Angostura.

[30] Figure 8 shows that an increase in crop area for the Angostura subbasin could exert a strong impact on its surface hydrology during winter, where higher reductions in base flow and soil moisture occurred. During the summer, the highly wooded areas of Paso Nacori and Oviachic are mostly affected by the agricultural extensification. In these areas, the increase in runoff is mostly driven by the reduction in ET and the increase in soil moisture. Comparing the changes in hydrological variables after modifying exclusively wooded areas (not shown) with those obtained by the experiment described above, we observe that base flow and soil moisture are least affected during the winter and monsoon seasons in Angostura than in any other subbasin. This could be due to the reduced presence of wooded areas in the domain. However, in Paso Nacori and Oviachic these changes produced different spatiotemporal responses for base flow, soil moisture, evapotranspiration and to a lesser extent runoff. During the monsoon, base flow, soil moisture, and evapotranspiration were reduced, while runoff increased. During winter, the opposite occurred, but runoff increased slightly. Consequently, the sensitivity of streamflow changes in Angostura depends on which current land use type is converted to cropland.

Figure 8.

Spatial distribution of water balance components for monsoon months (mn) and nonmonsoon months (nm) under current (C) and 10% altered (10%) conditions.

7. Conclusions

[31] We evaluated the sensitivity of streamflow in the major subbasins of YRB to land use change resulting from conversion of natural vegetation to unirrigated agriculture, and to the influence of midscale (NAM) and large-scale (El Niño–Southern Oscillation (ENSO) and Pacific Decadal Oscillation ± PDO)) climate regimes. Our analysis of the climatic variability of streamflow showed that during monsoon and winter streamflow differences vary according to the influence of the ENSO and PDO phenomena. Winter streamflow is most strongly affected by EN (much higher in EN than LN years) and in subbasins such as the Angostura, was higher than monsoon streamflow. Monsoon streamflow was influenced to a much lesser extent by ENSO (generally slightly lower in EN than LN years), however, it was also observed that monsoon streamflow in Paso Nacori decreased under the influence of LN during the cold PDO.

[32] Overall, the sensitivity of streamflow to climate variations dwarfed the sensitivity associated with land cover. For instance, the difference between winter flows in El Niño versus La Niña years was 76%, and for summer flows was 16%, in contrast to maximum monthly (winter) flow changes of 5% attributable to the largest land cover transition alternative tested. The highest sensitivity of runoff to agriculture extensification was found during winter. Furthermore, climate and land cover runoff sensitivities appear to be related; in El Niño (generally wet) years the sensitivity of streamflow to land cover change can be four times higher than in La Niña (generally dry) years. The sensitivity of streamflow to the PDO phase was smaller. Streamflow sensitivities to climate were highest when PDO and ENSO were in phase, e.g., during the 1970s, 1980s, and 1990s.

[33] In general, the positive anomaly of precipitation and streamflow increased the soil moisture content and reduced the soil moisture drought typical of the winter months. Wet soil moisture conditions in the subsurface favor evapotranspiration, producing soil moisture deficits that contribute to decreases in streamflow. Consequently, streamflow sensitivities to agricultural extensification in the YRB under different climatic conditions can be summarized by two mechanisms.

[34] The first mechanism results from the response of the spatial variability in streamflow to agricultural extensification. The predominant vegetation type under current conditions determines the magnitude and sign of the change in streamflow production. Specifically, under the dominant grass/shrub current conditions, an increase in crop surfaces results in a reduction in streamflow, as occurs in Angostura. Otherwise, streamflow increases due to the dominance of woodland in the current distribution of vegetation (Paso Nacori and Oviachic). The second mechanism results from the controls exerted by soil properties as evidenced during dry periods. This mechanism is related to the redistribution of subsurface moisture, and the controls of soil physics. In the model results, this is evidenced in the variability of base flow and soil moisture, and their role in reducing runoff generation in the northernmost portion of the YRB.

[35] The spatial distribution of vegetation and its changes affect water resources over the subbasins of the YRB to different extents. Given the importance of winter streamflows in refilling the water storages for agricultural activities, the sensitivities in Angostura are of special interest. On the other hand, further work is needed to explore the long-term impact of land use changes on soil moisture patterns and its impact on drought and flooding in the region.


[36] This publication was funded in part by the National Oceanic and Atmospheric Administration (NOAA) under grants NA04OAR4310078 and NA050AR4310014 to Duke University, by the Consejo Nacional de Ciencia y Tecnología, and by the NOAA Climate Program Office's Climate Prediction Program for the Americas under cooperative agreement NA060AR4310060 with the University of Washington, via the Joint Institute for the Study of the Atmosphere and Ocean (JISAO). This is JISAO contribution 1762.