1. Introduction
Almost 55 million acres of farmland is irrigated in the United States, accounting for more than 29 trillion gallons of water usage per year (NASS 2009). Most of this irrigation water is applied to the soil surface, creating an anthropogenic change to the land that impacts soil moisture but can ultimately influence clouds and precipitation through land–planetary boundary layer (PBL) coupling processes. The process chain by which soil moisture can impact clouds and precipitation [as defined by Santanello et al. (2011a)] involves various pathways of positive and negative feedbacks dependent on the relative sensitivities of 1) surface fluxes to soil moisture, 2) PBL evolution to surface fluxes, 3) entrainment fluxes at the top of the PBL to PBL evolution, and 4) the collective feedback of the atmosphere on the surface fluxes.
The local, direct impacts of irrigation on the surface flux–soil moisture component of this process chain are well understood. Modeling and observational studies agree that irrigation application reduces temperature and increases humidity via the repartitioning of latent heat flux and sensible heat flux (Moore and Rojstaczer 2002; Adegoke et al. 2003, 2007; Douglas et al. 2006; Bonfils and Lobell 2007; DeAngelis et al. 2010; Kueppers and Snyder 2012; Jiang et al. 2014). The ability of a numerical model to reproduce irrigation’s modifications to the surface energy balance is therefore essential for studies of land-use change impacts on climate (Zaitchik et al. 2005) and could potentially improve forecast skill in numerical weather prediction models (Ozdogan et al. 2010).
Irrigation-induced cloud and precipitation changes originate at the local scale and are regulated by feedbacks within the PBL. However, regional impacts of irrigation are more uncertain and vary by geographical area and climatological conditions. Past modeling studies have found irrigation can affect regional circulations (Chase et al. 1999; Lo and Famiglietti 2013; Huber et al. 2014) or induce remote precipitation responses (Im et al. 2014; Harding and Snyder 2012a,b), but others suggest that irrigation’s effects on surface climate are localized and do not extend very far into nonirrigated areas (Sorooshian et al. 2011). Further contributing to this regional-scale uncertainty is the fact that the atmospheric response produced by a regional simulation can be affected by differences in the details of irrigation representation, such as the timing and frequency of water application (Sorooshian et al. 2012). Irrigation’s regional impact could be dependent on the degree to which the land and atmosphere communicate changes to one another, as land–atmosphere (LA) coupling has been shown to influence precipitation patterns in studies of soil moisture feedbacks (Koster et al. 2002; Lawrence and Slingo 2005). Thus, a prerequisite to piecing together these regional impacts is an understanding of each irrigation method’s effect on land–PBL coupling and feedbacks at the local scale (i.e., the foundation of the process chain described above).
This paper seeks to address this research need by presenting a comprehensive assessment of irrigation’s impact on LA interactions using a high-resolution model test bed, multiple irrigation methods, and evaluation with a variety of surface observations. The purpose of this study is to 1) evaluate the sensitivity of a land surface model (LSM) multiyear spinup to several different irrigation methods and thresholds, 2) assess the impacts of irrigated spinups on regional coupled forecasts, 3) determine the effects of irrigation on land–PBL coupling, and 4) suggest recommendations for future irrigation implementation in LSMs. Previous studies and relevant background information are discussed in section 2, followed by a description of the model and data in section 3. Results from offline and coupled simulations, diagnosis of land–PBL coupling and evaluation with observations are discussed in section 4, with conclusions presented in section 5.
2. Background
a. Irrigation methods
The irrigation method chosen by a farmer is the product of numerous factors, including the associated monetary investment and labor intensity, the availability of water resources, and the topography of the landscape (S. Howser 2013, personal communication). In the central Great Plains states of Nebraska, Kansas, Iowa, and Missouri, center pivot sprinkler irrigation systems are by far the most widely utilized by farmers. As the method of choice on 68% of farms, sprinklers irrigate 80% of the total farmland acreage in this region (NASS 2009). Gravity systems, similar to flood irrigation, are inefficient from a water resources perspective, but are inexpensive, leading to their use on approximately 31% of farms. The most water-efficient method, drip irrigation systems, are costly and labor intensive and, as a result, are used on only 1% of farms in this area.
Recently, there has been a push for irrigation parameterizations that realistically reflect the variety and complexity of these irrigation practices. A popular representation of irrigation for regional applications forces soil moisture to saturation at a defined time interval in a manner similar to that of flood irrigation (Adegoke et al. 2003; Kueppers et al. 2007; Kueppers and Snyder 2012; Zaitchik et al. 2005; Jiang et al. 2014). Other parameterizations represent irrigation from an evapotranspiration (ET) or vapor flux perspective (Douglas et al. 2006; Segal et al. 1998; Evans and Zaitchik 2008), which best represents the water efficiencies of the drip irrigation method, or require soil moisture thresholds be exceeded before irrigation application occurs (Lobell et al. 2009; Qian et al. 2013; Tuinenburg et al. 2014). Ozdogan et al. (2010) simulated sprinkler irrigation by applying water as precipitation when the root-zone soil moisture fell below a triggering threshold. In some of the more sophisticated treatments of irrigation to date, Leng et al. (2014) used an offline land surface model to simulate the effects of irrigation on both surface fluxes and groundwater withdrawals while other studies have used an “irrigation demand factor” to prevent overirrigating the model soil surface (Pokhrel et al. 2012; Vahmani and Hogue 2014). The increasing complexity of these model parameterizations introduces the need to systematically assess their impacts on the LA interface.
The work presented in this paper is the first to comprehensively compare and evaluate drip, sprinkler, and flood irrigation parameterizations. As these methods are representative of the most common irrigation parameterizations in use today, this work should provide value to a range of future irrigation studies.
b. Coupled impacts of irrigation
Modeling studies utilizing irrigation parameterizations have shown that irrigation can significantly impact the surface energy balance and near-surface temperature from local to global scales. Boucher et al. (2004) estimated that irrigation produces a radiative forcing in the range of 0.03–0.1 W m−2 at the global scale because of the additional atmospheric water vapor, and it induces surface cooling of up to 0.8 K over irrigated land areas. Using the Community Atmosphere Model, version 3.3 (CAM3.3), Lobell et al. (2009) concluded that irrigation significantly reduced the model’s warm bias over several heavily irrigated areas in the United States, such as California and Nebraska. In a study using the Noah LSM (Chen et al. 1996) over the continental United States, Ozdogan et al. (2010) found increases in surface latent heat flux of up to 100 W m−2 in California, as well as other regions across the United States, when simulating sprinkler irrigation. Similarly, in a study using the Regional Atmospheric Modeling System (RAMS), Adegoke et al. (2007) found irrigation increased latent heat flux by 36% in Nebraska, decreasing temperatures by 1.2°C and increasing dewpoints by 2.3°C. These cross-scale modeling results are corroborated by observational studies that have found decreases in surface temperature correlated with increasing spatial extent of irrigated agriculture in California (Bonfils and Lobell 2007) and the high plains aquifer of the southern Great Plains (Mahmood et al. 2013).
The potential ability of irrigation to mitigate or reinforce wet and dry periods through its impact on precipitation has important implications, as these climatological extremes can influence farmers’ yields. Observational datasets have shown an enhancement of precipitation downwind of heavily irrigated areas in the Great Plains and Texas high plains (DeAngelis et al. 2010; Moore and Rojstaczer 2002), but irrigation-enhanced precipitation can actually lead to a net water loss as recycled precipitation often falls away from the source and is outweighed by ET increases (Harding and Snyder 2012a,b; Wei et al. 2013). Simulating the effects of irrigation on precipitation amounts and patterns often yields results that are dependent on the model used (Tuinenburg et al. 2014), the antecedent soil moisture conditions (Harding and Snyder 2012b), or the region of interest (Kueppers et al. 2008). Irrigation has even shown the potential to weaken the Great Plains low-level jet, thus increasing July precipitation by almost 50% downwind of irrigated areas (Huber et al. 2014).
Despite the importance of LA coupling and PBL evolution in the soil moisture–precipitation process chain, it can be seen from the aforementioned studies that much of the irrigation research to date either restricts analysis to surface flux–soil moisture interactions or is motivated to discern precipitation impacts with little regard for the PBL feedbacks that connect soil moisture to precipitation. An exception is a recent study by Qian et al. (2013), which explored the impacts of irrigation on the diurnal cycle of near-surface fluxes, temperature, clouds, and precipitation in the southern Great Plains using the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2005). Their study showed irrigation repartitions surface fluxes and also increases the probability of shallow clouds by decreasing the lifting condensation level (LCL) more than the PBL height (PBLH).
The work presented in this paper takes an even more comprehensive approach by using an irrigated LSM spinup to capture soil moisture anomalies and by evaluating the model sensitivity to irrigation algorithms that differ in frequency, timing, and application. Furthermore, this work diagnoses the LA coupling and feedbacks using a high-resolution modeling environment forced by best available surface and satellite observations and utilizes the observational datasets necessary to evaluate irrigation schemes and coupled impacts.
3. Methods
a. Model and experimental design
This study utilizes NASA’s Land Information System (LIS; Kumar et al. 2006), version 6.1, to complete offline spinups and generate initial conditions for coupled forecasts with the Advanced Research version of the WRF (ARW). LIS is a flexible land surface modeling and data assimilation framework that allows users to choose from a variety of LSMs that are forced and constrained by best available surface and remote sensing observations. The ARW is a community mesoscale research model with an Eulerian mass solver, a terrain-following vertical coordinate system, and multiple physics options (Skamarock et al. 2005). LIS has been fully coupled to WRF under the NASA Unified WRF (NU-WRF; Peters-Lidard et al. 2015) framework. This model configuration allows for long-term offline land surface model spinups using observed atmospheric forcing and creates a better representation of LA interactions through the ability to characterize the land surface at the same spatial scales as cloud and precipitation processes (Kumar et al. 2006). In this way, the LIS–WRF configuration has shown skill in studies in which soil moisture anomalies and LA interactions play a prominent role (Santanello et al. 2009, 2011a, 2013b).
The study area is a 500 km × 600 km region of the central Great Plains including portions of Nebraska, Kansas, Iowa, and Missouri, shown in Fig. 1. This area provides a steep irrigation gradient, as the western region is heavily irrigated, but minimal irrigation occurs in the eastern section. Present in the domain are three flux observation sites from which data will be used to evaluate model results, discussed more in section 3c. The Noah LSM, version 3.2, was run offline (uncoupled) within the LIS framework at 1-km resolution for 5 years (2005–10) using four different irrigation schemes, discussed further in section 3b, and a control run (no irrigation; hereafter Control). In addition to offline sensitivity experiments, each LIS–Noah spinup was used to initialize a 2-day WRF forecast to study the relative impacts of each irrigation method on the PBL evolution and regional weather forecast. Two simulation periods were chosen, one in a wetter-than-normal year and one in a drier-than-normal year, to evaluate the sensitivity of the model to the background climate conditions. The wet and dry years were determined from the domain-averaged Phase 2 of the North American Land Data Assimilation System (NLDAS-2) soil moisture data, which showed 2006 to have the lowest and 2008 to have the highest soil moisture over the 5-yr spinup period. The LIS–WRF configuration was run for 48 h at 1-km resolution on 30–31 July 2006 (dry) and 25–26 May 2008 (wet). These 2-day periods exhibited the driest and wettest soil moisture in the NLDAS-2 forcing data in the dry and wet year, respectively, during the irrigation season.
LSM and coupled simulations were run in a single domain in the central Great Plains of the United States, denoted by the yellow box in the inset. Simulation domain has green dots to denote grid cells classified as “irrigated” according to USGS land-cover data. Stars mark the sites used for analysis, including irrigated and rainfed sites in Mead, Nebraska (blue); ARM-SGP site E4 in Plevna, Kansas (orange); and point A (purple).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
LSM and coupled simulations were run in a single domain in the central Great Plains of the United States, denoted by the yellow box in the inset. Simulation domain has green dots to denote grid cells classified as “irrigated” according to USGS land-cover data. Stars mark the sites used for analysis, including irrigated and rainfed sites in Mead, Nebraska (blue); ARM-SGP site E4 in Plevna, Kansas (orange); and point A (purple).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
LSM and coupled simulations were run in a single domain in the central Great Plains of the United States, denoted by the yellow box in the inset. Simulation domain has green dots to denote grid cells classified as “irrigated” according to USGS land-cover data. Stars mark the sites used for analysis, including irrigated and rainfed sites in Mead, Nebraska (blue); ARM-SGP site E4 in Plevna, Kansas (orange); and point A (purple).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
The LIS–WRF simulations were completed using a single domain with 43 vertical levels and a time step of 5 s. The Monin–Obukhov surface-layer scheme as well as Goddard microphysics and short- and longwave radiation were utilized. To allow the model to explicitly resolve convection at 1-km resolution, a cumulus parameterization was not used. Initial and boundary conditions were provided by the North American Regional Reanalysis (NARR; Mesinger et al. 2006) at 3-hourly intervals. The Mellor–Yamada–Janjić (MYJ; Janjić 1994) PBL scheme was used for this study, which exhibits oscillatory PBLH estimations in the daytime hours based on TKE. Thus, in analysis of the model output, PBLH is estimated using a bulk Richardson approach, in which the pressure corresponding to the first model level where the bulk Richardson number exceeds 0.25 is assumed to be the top of the PBL (Sivaraman et al. 2013). An additional NU-WRF run without a spinup (hereafter referred to as NoSpin) was completed for each case to demonstrate the impact of initializing from an LIS spinup versus that of the coarse atmospheric analysis (i.e., NARR) data.
To ensure the land surface states were fully equilibrated by the desired initialization time in 2006, a second spinup from 2003 to 2010 was completed and compared to the original. The 2005 spinup reached equilibrium in less than a year for soil moisture for the top three layers and for all layers with regards to soil temperature. Fourth-layer (bottom) soil moisture equilibrated by mid-2007. However, differences in fourth-layer soil moisture in July 2006 are small (less than 0.02 m3 m−3) and negligible for the purposes of this study. Therefore, we determined the model to be appropriately spun up by the July 2006 WRF initialization time.
b. Irrigation parameterizations
This study uses three irrigation schemes—flood, sprinkler, and drip—each differing in the frequency, type, and timing of water application. These methods are implemented in the Noah LSM within the LIS framework.
The sprinkler method (hereafter referred to as Sprinkler) is derived from Ozdogan et al. (2010) but has been modified to employ a user-specified irrigation rate as opposed to one based on crop water demand. In this way, the Sprinkler scheme used here closely represents the farmer’s perspective of water application and application rates. Water is applied uniformly as precipitation at a rate of 5 mm h−1 when the root-zone moisture availability (RZMA) falls to 10% above the stress point. The irrigation water shuts off when the RZMA reaches 80% of the maximum soil moisture.
The flood (hereafter Flood) method builds on Evans and Zaitchik (2008) by adding a threshold used to evaluate the soil moisture state before triggering irrigation. This method applies water to the root zone at 0900 local time (LT) if the root-zone soil moisture falls below a threshold with respect to the wilting point (Yilmaz et al. 2014). Water is then applied until the top layer is saturated and saturation is sustained for 30 min. To evaluate the model sensitivity to this threshold, two Flood irrigation spinups were completed: one using a threshold of 25% above the wilting point and the other at 75% (hereafter Flood25 and Flood75, respectively).
The drip (hereafter Drip) method originates from Evans and Zaitchik (2008) and is designed to provide an optimal amount of water—enough to allow for transpiration without stress but without any excess water application. Soil moisture impacts ET in the Noah land surface model through the canopy resistance. The Drip parameterization calculates the canopy resistance and resultant ET twice—once using the current soil moisture and the second time assuming no soil moisture stress. The unstressed transpiration value is used in the simulation and the difference between the two values is the irrigation requirement to avoid stress. This approach results in soil moisture being largely unchanged during the simulation, as the water required is immediately transpired rather than added to the soil column.
The irrigation algorithms are applied homogeneously to each 1-km grid cell exhibiting an irrigable land-use classification. In this study, the 24-class category U.S. Geological Survey (USGS) land-use classification data were used within LIS. This dataset contains two irrigable categories: irrigated cropland and pasture (fully irrigated) and mixed dryland–irrigated cropland and pasture (partially rain fed). The Drip and Sprinkler methods do not distinguish between these categories, but the Flood algorithm uses half (all) of the maximum soil moisture content in calculating the irrigation water requirement for partially rainfed (fully irrigated) grid cells. However, of the two irrigable land-use categories, only fully irrigated grid cells appear in our domain. An additional important criterion needed to activate the methods ensures that it is irrigation season by requiring that the gridcell greenness vegetation fraction (GVF) exceed 40% of the climatological annual range of GVF, after Ozdogan et al. (2010). In LIS–Noah, GVF is derived from satellite-based monthly climatology data also at 1-km resolution.
c. Evaluation
Several techniques and observational datasets were employed to evaluate the offline and coupled model output. The Land Verification Toolkit (LVT; Kumar et al. 2012) is a software tool designed to enable robust evaluation of LIS output against observations from a variety of sources. This study employs the LVT for analysis of LIS–Noah output against observations of fluxes, soil moisture, and soil temperature from the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement Program in the southern Great Plains (ARM-SGP). In addition, the Model Evaluation Tools (MET) software is used to compare the coupled LIS–WRF output to point observations within the study area (Developmental Testbed Center 2013). MET was developed at the National Center for Atmospheric Research (NCAR) through grants from the U.S. Air Force Weather Agency (AFWA) and the National Oceanic and Atmospheric Administration (MET user’s guide). Additional point data were provided by the AmeriFlux group of the DOE’s Oak Ridge National Laboratory and were analyzed to create average daily cycles of fluxes to verify those simulated by LIS–Noah.
LA coupling is examined through the use of local-scale land–atmosphere coupling diagnostics (LoCo; Santanello et al. 2011b), including mixing diagrams (MDs; Santanello et al. 2009, 2011a, 2013b), PBL–evaporative fraction (EF) analyses (Santanello et al. 2009), and the concept of the LCL deficit (Santanello et al. 2011a, 2013b). These tools have proven useful in determining the impact of soil moisture perturbations on surface forcing and the subsequent PBL response as well as the sensitivity of WRF to various LSM and PBL combinations. This makes LoCo diagnostics ideal for intercomparison of the impacts of various irrigation schemes as well.
Mixing diagrams represent the diurnal evolution of near-surface humidity and potential temperature using vectors in energy space, allowing for the quantification of heat and moisture budgets in the PBL and several related metrics (e.g., Bowen and entrainment ratios). These diagrams are constructed using the 2-m temperature and humidity at a particular point converted to heat and moisture energy space via multiplication by the specific heat of water and the latent heat of vaporization, respectively. The resultant values for each daytime hour are plotted, creating the solid line shown in the mixing diagrams. The dashed lines are vectors, which represent the fluxes of heat and moisture from the surface and atmosphere. In this analysis, we treat the residual vector of the mixing diagrams as the atmospheric response vector (Santanello et al. 2013a), which is typically dominated by entrainment fluxes but also includes horizontal advection. The slope of the surface (atmospheric response) vector is exactly equal to the surface (entrainment) Bowen ratio, and the magnitude of the vector components are proportional to the surface (entrainment) fluxes of heat, given by the y component, and moisture, given by the x component. For a more comprehensive discussion of mixing diagram theory and LoCo diagnostics, interested readers are referred to Santanello et al. (2009) and Betts (1992).
4. Results
a. Offline spinup
1) Regional and multiyear impacts
Figure 2 shows monthly, domain-averaged differences from Control in top-layer soil moisture SM, latent heat flux Qle, and sensible heat flux Qh for each of the irrigation methods during the 5-yr spinup. The seasonal cycle of irrigation application is evident, with peaks in midyear and decreases toward the end of each year. Of particular note is the memory of the soil to the previous growing season’s irrigation practices, shown in SM increases that linger through the winter season. However, this residual SM anomaly has only a negligible impact on winter fluxes. The 25% threshold imposed on the Flood irrigation method is more restrictive than the 75% case, requiring the soil dry to a greater degree before irrigation will be triggered. Thus, Flood75 results in greater increases in soil moisture than Flood25, while Sprinkler irrigation shows the largest changes of all methods. As anticipated, Drip exhibits zero changes to soil moisture content because of the nature of the algorithm, as additional water is immediately used for transpiration.
Domain-averaged monthly change from Control during the 5-yr LIS–Noah spinup for top-layer (upper 0–10 cm) (top) SM, (middle) Qle, and (bottom) Qh.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Domain-averaged monthly change from Control during the 5-yr LIS–Noah spinup for top-layer (upper 0–10 cm) (top) SM, (middle) Qle, and (bottom) Qh.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Domain-averaged monthly change from Control during the 5-yr LIS–Noah spinup for top-layer (upper 0–10 cm) (top) SM, (middle) Qle, and (bottom) Qh.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
The increased soil moisture repartitions the surface fluxes in the Sprinkler and Flood runs, consistent with previous modeling studies’ findings of the impact of SM on fluxes (Adegoke et al. 2007; Qian et al. 2013). Flood75 and Sprinkler increase latent heat flux by up to 7.0 and 8.0 W m−2, respectively. Although no soil moisture changes are noted in the Drip method, the ET modification causes latent heat flux to rise by as much as 3.5 W m−2. Sensible heat flux decreases by a complimentary amount for each of the methods. The greatest changes to the energy balance during the 5-yr spinup occur in 2006, noted as a dry regime for the study area, while the wetter regimes of 2007 and 2008 exhibit the smallest changes. Such results are expected as dry periods are characterized by a lack of precipitation, low soil moisture values, greater plant stress, and decreased evapotranspiration—conditions that will trigger the irrigation algorithms to turn on more frequently than in a wet or normal regime.
Irrigated grid cells are most commonly found in the western third of the domain and account for only 4% of the total study area. Thus, impacts are minimized in the previous analyses because of the heavy weight of nonirrigated grid cells and the averaging of the model output twice—once temporally and a second time spatially with a majority of nonirrigated areas. The following section presents analyses for a dry (2006) and wet (2008) year spatially to determine the variation of impacts within the domain during contrasting antecedent soil moisture conditions.
2) Spinup results in a dry and wet regime
Irrigation impacts vary across the study area and are sensitive to time and method. Seasonally averaged changes in SM for the dry and wet regimes are presented in Figs. 3 and 4, respectively. In this case, “seasonal” refers to the irrigation season, defined as from 1 May to 30 September, as this is the primary growing season in Nebraska (Adegoke et al. 2003). Sprinkler irrigation impacts SM and fluxes the most, increasing SM by 0.16–0.2 m3 m−3 and latent heat flux by at least 100 W m−2 over most irrigated grid cells. Increases in latent heat flux (Figs. 5, 6) as a result of the Drip algorithm slightly exceed those due to SM effects in Flood25 with increases of up to 85 and 50 W m−2 for the two methods, respectively. Complementary decreases in sensible heat flux are apparent in these irrigated areas. The effects of irrigation are muted when the background precipitation is greater, as the impacts of each method are consistent with those during the dry regime, but are smaller in magnitude. This is especially noticeable near the Nebraska–Kansas border (western part of the domain near 40°N), where irrigation generally reduces sensible heat flux by 40–50 W m−2 more in the dry regime as compared to the wet. The energy balance is consistently impacted the most in the southwestern part of the domain, regardless of method or regime. This feature could potentially indicate that the soil in this region dries out more quickly than over the other irrigated areas, or that this area consistently experiences less precipitation in relation to the rest of the domain during these two regimes.
Seasonally averaged (May–September) change in top-layer SM content during the dry regime of 2006 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Seasonally averaged (May–September) change in top-layer SM content during the dry regime of 2006 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Seasonally averaged (May–September) change in top-layer SM content during the dry regime of 2006 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for the wet regime of 2008.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for the wet regime of 2008.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for the wet regime of 2008.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 3, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 4, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 4, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 4, but for Qle.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
LIS–Noah spinups provide initial conditions for the WRF runs in the form of soil moisture and soil temperature. It is important to note that although soil moisture is not impacted directly by Drip irrigation, the method does have an effect on soil temperature. Latent heat flux increases caused by the Drip algorithm impact the land surface energy balance calculated by the Noah LSM, which simultaneously solves for fluxes and surface temperature. Thus, initial conditions for soil temperature are impacted by the Drip method via changes to the latent heat flux and surface temperature and are of similar magnitude to impacts seen via Flood25.
b. Coupled results
Figures 7 and 8 present the change from Control in the midday 2-m temperature T2 and humidity Q2, respectively, for each of the irrigation methods in the coupled LIS–WRF simulations. Quiescent synoptic conditions in the dry regime forecast period amplify the impact of the irrigation-induced soil moisture perturbations in the coupled run forecast. Evidence of this is apparent in the comparison of Drip, which exhibits no direct soil moisture changes and therefore only small forecast changes, to Sprinkler, Flood75, and Flood25, which exhibit both direct and indirect impacts. Increased soil moisture and latent heat flux at the surface cause sensible heat flux and temperature to drop while increasing humidity over irrigated areas in the coupled run. Midday T2 decreases by as much as 4 K in the Sprinkler and Flood75 runs over irrigated grid cells while water vapor mixing ratio increases by up to 4 g kg−1.
Difference from Control in LIS–WRF-simulated T2 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods at 1400 LT 31 Jul 2006 (dry regime). All methods decrease T2 over irrigated grid cells, ranging from a slight reduction (Drip) to almost 5 K (Sprinkler).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Difference from Control in LIS–WRF-simulated T2 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods at 1400 LT 31 Jul 2006 (dry regime). All methods decrease T2 over irrigated grid cells, ranging from a slight reduction (Drip) to almost 5 K (Sprinkler).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Difference from Control in LIS–WRF-simulated T2 using (a) Flood25, (b) Flood75, (c) Drip, and (d) Sprinkler irrigation methods at 1400 LT 31 Jul 2006 (dry regime). All methods decrease T2 over irrigated grid cells, ranging from a slight reduction (Drip) to almost 5 K (Sprinkler).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 7, but for Q2.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 7, but for Q2.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 7, but for Q2.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Although the greatest changes to soil moisture and fluxes occurred in the southwestern part of the domain in the spinup results, the greatest impacts in the coupled run appear just north of the Nebraska–Kansas border (near 40°N). Southerly winds advect the irrigation-cooled and moistened air northward, reducing the temperature downwind by 1–2 K and increasing water vapor by 1.5 g kg−1. Thus, the maximum impact in southern Nebraska is likely due to a combination of the direct impacts, resulting from the densely irrigated area, with the indirect effects stemming from its location downwind of other irrigated grid cells.
The wet regime simulation period featured precipitation events associated with a cold frontal passage, leading to a greater impact of the initial conditions on cloud and precipitation patterns than on surface states. Temperature is still reduced by 1–2 K and humidity increases by up to 2 g kg−1 near the Nebraska–Kansas border, but overall, the direct surface impacts are muted as compared to the dry regime. Irrigation causes precipitation to vary in timing and location from Control, but changes to the magnitude of total accumulated precipitation over the simulation period are small (Fig. 9). Sprinkler increases rainfall by only 2.6%, while Drip creates a 1.8% reduction in precipitation.
As in Fig. 7, but for total accumulated precipitation over the 2-day LIS–WRF simulation on 25–26 May 2008 (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 7, but for total accumulated precipitation over the 2-day LIS–WRF simulation on 25–26 May 2008 (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
As in Fig. 7, but for total accumulated precipitation over the 2-day LIS–WRF simulation on 25–26 May 2008 (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
c. LoCo diagnostics
1) Mixing diagrams and EF–PBLH
Figure 10 presents mixing diagrams illustrating the diurnal change in temperature and humidity for each of the experiments at an irrigated grid cell (Fig. 1, point A). Point A was chosen as an analysis site as it has an irrigated land-cover classification and interannual soil moisture characteristics that are representative of the dry and wet regimes. At this site, the NoSpin run is the driest as compared to the spinup simulations, which were run with observed forcing, consistent with the idea of a dry bias in the warm regimes in the Noah LSM (Chen et al. 2007). The “shepherd’s hook” appearance of the mixing diagram evolution is indicative of a moistening of the PBL through strong surface evaporation in the morning (staff–handle portion) followed by drying due to PBL growth and entrainment in the afternoon (bowed top). This evolution is consistent with previous results of LoCo analysis during dry regimes (Santanello et al. 2009).
MDs for the LIS–WRF simulations during the (a) dry regime and (b) wet regime at point A. The line representing the Control simulation is boldface. The solid lines are T2 and Q2 plotted in energy space from 0700 to 1700 LT. The tail of the surface vector begins at the initial time point (bottom of solid line) and the residual vector has its tail at the final time point above. The entrainment ratio of heat Ah gives the proportion of sensible heat input to the PBL by entrainment Hent compared to that of the surface Hsfc, and similarly for latent heat input Ale. (c) Daytime mean evaporative fraction vs daily max PBLH at point A on 31 Jul (dry regime) and (d) 25 May (wet regime). Each simulation is represented by a marker of a particular color and style. Note there is overlap in the markers in both the dry and wet regimes.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
MDs for the LIS–WRF simulations during the (a) dry regime and (b) wet regime at point A. The line representing the Control simulation is boldface. The solid lines are T2 and Q2 plotted in energy space from 0700 to 1700 LT. The tail of the surface vector begins at the initial time point (bottom of solid line) and the residual vector has its tail at the final time point above. The entrainment ratio of heat Ah gives the proportion of sensible heat input to the PBL by entrainment Hent compared to that of the surface Hsfc, and similarly for latent heat input Ale. (c) Daytime mean evaporative fraction vs daily max PBLH at point A on 31 Jul (dry regime) and (d) 25 May (wet regime). Each simulation is represented by a marker of a particular color and style. Note there is overlap in the markers in both the dry and wet regimes.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
MDs for the LIS–WRF simulations during the (a) dry regime and (b) wet regime at point A. The line representing the Control simulation is boldface. The solid lines are T2 and Q2 plotted in energy space from 0700 to 1700 LT. The tail of the surface vector begins at the initial time point (bottom of solid line) and the residual vector has its tail at the final time point above. The entrainment ratio of heat Ah gives the proportion of sensible heat input to the PBL by entrainment Hent compared to that of the surface Hsfc, and similarly for latent heat input Ale. (c) Daytime mean evaporative fraction vs daily max PBLH at point A on 31 Jul (dry regime) and (d) 25 May (wet regime). Each simulation is represented by a marker of a particular color and style. Note there is overlap in the markers in both the dry and wet regimes.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
The Control and Drip irrigation simulations exhibit almost the same diurnal evolution of temperature and humidity, suggesting that changes to the SM initialization are required to significantly impact T2, Q2, and fluxes. In these runs, the surface sensible heat flux is greatest, leading to the most PBL growth and dry air entrainment and a PBL that is mostly balanced between surface and atmospheric/entrainment inputs. In contrast, the surface provides energy to the atmosphere predominately in the form of latent heat for the Flood75, Sprinkler, and Flood25 runs. In fact, negative Bowen ratios in the Flood75 and Sprinkler runs are a result of negative sensible heat flux, indicating heat loss from the atmosphere to the surface—a common microclimate situation over irrigated fields, referred to as the “oasis effect” (Oke 1978) or “sensible heat advection” (Brakke et al. 1978). With the exception of Drip, the surface sensible heat input is reduced to such an extent in the irrigated runs that the entrainment flux of sensible heat, even though reduced because of slower PBL growth, becomes the dominant source of heat input to the PBL.
The impact of irrigation on the PBL budget is muted in the wet year. Qualitatively, these mixing diagrams indicate strong surface moisture input and a much smaller range in temperature and humidity as compared to the dry year. This is consistent with the results of Santanello et al. (2013a), where the impacts of different LSM choices were maximized during dry regimes and much less during wet. Flood75 and Sprinkler methods result in more evaporation, but these impacts are not translated downstream as the changes to T2, Q2, fluxes, and entrainment are negligible.
Daily maximum PBLH and daytime mean EF [defined as the ratio of Qle to (Qle + Qh)] are integrative metrics of the land surface and PBL and thus allow evaluation of the PBL response to changes in surface forcing. Plots of PBLH versus EF are shown in Figs. 10c and 10d for the dry and wet year, respectively. In the dry year, EF is proportional to the amount of water applied by the irrigation methods (e.g., EF > 1 for Flood75 and Sprinkler but nearly unchanged for Drip), while the PBLH decreases in response to this surface forcing. The relatively dry surfaces of the Control and Drip runs (low EF) force the PBL to grow rapidly, ultimately reaching more than 2 km. Flood25 increases EF by about 0.5, reducing PBLH by a few hundred meters. However, the greatest changes occur with Flood75 and Sprinkler, as these methods reduce PBLH by almost 1 km at this location. This analysis makes apparent the sensitivity of the PBL response to irrigation method and the range of impacts of each scheme in the dry year.
Wet regime impacts to PBLH and EF indicate that irrigation only minimally impacts the PBL at this site. The effect of irrigation on EF is dependent on method, with small increases for the most water intensive methods (Flood75 and Sprinkler), but the differences in EF are quite small compared to the dry regime. As a result, PBLH is not affected, as maximum height remains around 1.4 km for all runs.
2) LCL deficit
Essential to the development of convective clouds is the requirement of the PBLH to exceed the LCL. A comparison of the PBLH and LCL evolution on diurnal time scales, referred to as LCL deficit (Santanello et al. 2011a), is analyzed at point A as well as spatially over the domain. A negative LCL deficit reveals that the PBLH (millibars) has exceeded the LCL, therefore indicating the potential for cloud development. Figure 11 shows a time series of the LCL deficit at point A for the second day of the dry regime coupled runs. In this case, the LCL deficit never becomes negative, but Sprinkler and Flood75 steadily decrease the LCL deficit throughout the morning hours (during the moistening and PBL growth phase seen in the mixing diagrams), allowing both to approach zero around 1000 LT. As the day progresses, the LCL rises faster than the PBL, resulting in an increasingly positive LCL deficit for all simulations.
Hourly LCL deficit at point A on 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Hourly LCL deficit at point A on 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Hourly LCL deficit at point A on 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Although the LCL deficit stays positive at this site, there are other regions of the domain where irrigation creates a negative LCL deficit in the morning. At 1000 LT, Sprinkler irrigation reduces the LCL deficit by about 60 mb over irrigated areas in the western portion of the domain (Fig. 12, top). This decrease is large enough to make the LCL deficit negative over those grid cells. Both PBLH and LCL are reduced (higher pressure), but the LCL reduction outweighs that of PBLH, thus driving the LCL deficit decrease. Another notable feature is the advection of moistened and cooled air northward that lowers the LCL in regions downstream of irrigated grid cells (as a function of the advection of the impacts on T2 and Q2). However, toward early evening (1700 and 1800 LT), the LCL deficit increases over the heavily irrigated areas as the PBL breaks down sooner than in the Control run (Fig. 12, bottom).
Change from Control in LCL deficit using the Sprinkler method at (top) 1000 and (bottom) 1800 LT 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Change from Control in LCL deficit using the Sprinkler method at (top) 1000 and (bottom) 1800 LT 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Change from Control in LCL deficit using the Sprinkler method at (top) 1000 and (bottom) 1800 LT 31 Jul (dry regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Although irrigation in Flood75 and Sprinkler create a negative LCL deficit in the morning, these changes are not reflected in the cloud field, as cloud development is not simulated until 1400 LT. Thus, irrigation moves the PBL toward a more saturated state in the morning, increasing the proclivity for clouds, but the dry conditions are extreme enough to prevent cloud formation.
At point A in the wet year, the LCL deficit is strongly negative for all methods, and although it increases steadily after 0900 LT, it remains negative throughout the day. As compared to Control, the irrigation methods only slightly reduce the LCL deficit between 1000 and 1300 LT. Spatially, the wet year exhibits large areas of sustained negative LCL deficit values that agree with the location of clouds indicated by the cloud water mixing ratio, shown in Fig. 13. Once again, the LCL deficit is reduced over the irrigated areas, but synoptic rather than surface forcing is the catalyst for clouds and precipitation on this day.
(top) The LCL deficit and (bottom) vertically integrated cloud water mixing ratio for the LIS–WRF Control simulation at 0800 LT 25 May (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
(top) The LCL deficit and (bottom) vertically integrated cloud water mixing ratio for the LIS–WRF Control simulation at 0800 LT 25 May (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
(top) The LCL deficit and (bottom) vertically integrated cloud water mixing ratio for the LIS–WRF Control simulation at 0800 LT 25 May (wet regime).
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Synthesizing the results from the dry and wet year indicates that cloud development requires a strongly negative LCL deficit over many hours. Furthermore, in the absence of larger-scale forcing, as in the dry year, irrigation results in both positive and negative feedbacks on the PBL depending on the time of day. In the morning when the PBL and LCL are shallow, the irrigation perturbation decreases the LCL more than PBLH, consistent with the results of Qian et al. (2013), thereby increasing the chance for convective cloud development (positive feedback). However, the integrative nature of the PBL is such that the memory of reduced heating over the course of the day in irrigated areas causes the PBL to collapse sooner, reducing the chance of convective cloud development in the late afternoon (negative feedback). Advection of moistened and cooled air northward lowers the LCL downwind of the irrigated areas while minimally impacting PBLH, thereby creating a positive feedback downwind that is present throughout the day. Analyses that do not consider the diurnal cycle or changes to background conditions from day to day are likely to average across these feedback mechanisms.
d. Evaluation against observations
Because of the inherent human and plot-scale influences on irrigation practices, evaluation of irrigation physics in models is not as straightforward as traditional model validation of thermodynamic states or fluxes. Here, we survey at first order an array of potential validation approaches ranging from point to satellite-retrieved scales.
Observations from the ARM-SGP site E4 in Plevna, Kansas, were used to assess the offline and coupled simulations using the Land Verification Toolkit and LoCo diagnostics, respectively. The land-use category for the grid cell representing the E4 location in the model is not irrigable, thereby limiting the comparisons to only the Control and NoSpin simulations. LIS–Noah underestimates the July 2006 average daily cycle of Qle by more than 50 W m−2 in the afternoon followed by a smaller overestimation during the nighttime hours. As a result, Qh is overestimated in the afternoon and underestimated during the early morning hours, causing soil temperature to fluctuate more over the daily cycle than what is observed.
Average daily cycles generated from AmeriFlux towers at a rainfed and an irrigated site in Mead, Nebraska, reveal that LIS–Noah overestimates Qh and underestimates Qle at each location, in a way similar to that at the E4 site. The model grid cell associated with the irrigated Mead site is not of an irrigable land use, again preventing an intercomparison of the irrigation methods, but the observations expose some insight into the microclimate at these sites. Of particular note is the negative sensible heat flux in the observations associated with the oasis effect—the same effect noted at Point A with Flood75 and Sprinkler methods. Thus, it is possible that the simulation of the fluxes at this site would have been improved if the land use were irrigable, allowing the irrigation methods, especially Flood75 or Sprinkler, to activate here.
For a more robust evaluation against observations, the MET toolkit was used to match point observations within the domain to the complementary model grid cell (together called a pair) and to generate statistics assessing the skill of the coupled LIS–WRF runs. In the study area, the number of available observations, and therefore pairs, varies between 60 and 80 depending on the hour. It should be noted that only one of these pairs is located at a grid cell of irrigable land-cover classification. Thus, any differences in statistics revealed between the Control and irrigation methods are mostly a result of the indirect impacts of irrigation.
The most noteworthy feature apparent in the statistical analysis is the fact that the coupled runs initialized from any of the LIS spinups consistently show reduced daytime RMSE (Fig. 14) and MAE for temperature and humidity in the dry year. In addition, Sprinkler irrigation reduces the daytime warm temperature bias the most and improves the dry bias in Q2. In the wet year, model bias is initially reduced in the first 6 h by the spinups, but the precipitation events thereafter confound the statistics. The impact of simply including any type of spinup is generally dominant over any individual irrigation scheme effects because of the nonirrigated land-cover classification associated with most of the pairs. Overall, the Plevna, Mead, and MET analyses highlight the importance of using an LSM spinup, as well as the difficulties inherent in the evaluation of irrigation methods with point observations.
Hourly RMSE for (top) T2 and (bottom) Q2 for each of the LIS–WRF and the NoSpin simulations generated by the MET for the dry regime.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Hourly RMSE for (top) T2 and (bottom) Q2 for each of the LIS–WRF and the NoSpin simulations generated by the MET for the dry regime.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Hourly RMSE for (top) T2 and (bottom) Q2 for each of the LIS–WRF and the NoSpin simulations generated by the MET for the dry regime.
Citation: Journal of Hydrometeorology 16, 3; 10.1175/JHM-D-14-0203.1
Observation-based analysis of fluxes within irrigated areas was challenging because of the problem of evaluating short model simulations against limited observations when the exact timing of real world irrigation applications is unknown. Nevertheless, we were able to confirm the realism of the USGS irrigated areas map by comparison with county-level freshwater withdrawal data from a USGS report on U.S. water use (Hutson et al. 2004). The report confirms the most heavily irrigated areas are located in the western part of the domain, as is the case in the land-use classification data. We also compared the USGS product to the MODIS-derived, contiguous U.S. irrigation map of Ozdogan and Gutman (2008) and found reasonable agreement across the simulation domain (not shown), but that the spatial extent of irrigated area in the USGS land cover is less than the MODIS-derived dataset. The USGS data give an irrigated area of 12 341 km2 for the entire study area, but it is estimated by the Census of Agricultural Farm and Ranch Irrigation Survey that 15 505 km2 are irrigated in Nebraska alone (NASS 2009).
Evapotranspiration as a proxy for irrigation is a promising avenue for validation as irrigated regions exhibit a markedly different ET signature than surrounding areas during dry years. The general pattern of ET produced by LIS–Noah is comparable to that of the MODIS–surface radiation budget (SRB; Tang et al. 2009) ET product from the University of Washington. However, a single MODIS overpass per day is not capable of reproducing the dynamics influencing the model output, and the dates of irrigation in the model do not necessarily match the exact dates of irrigation in the real world, thereby ruling out a rigorous validation of absolute ET magnitudes. The planned spatial expansion of the University of Nebraska–Lincoln’s products using the Mapping Evapotranspiration at High Resolution and Internalized Calibration (METRIC; Irmak et al. 2011) technique will likely make it a valuable high-resolution spatial dataset for validation of ET and irrigation in future studies. Although these spatial and satellite-derived products are currently limited, by and large, they show more potential for future validation of irrigation schemes at the regional scale.
5. Discussion and conclusions
This study has used a high-resolution model test bed and several irrigation parameterizations and thresholds to assess the impact of irrigation on LA interactions during a dry and a wet regime. Irrigation’s ability to mitigate the soil moisture stress imposed by dry regimes directly impacts the surface energy budget, PBL growth, and ambient weather. The extent to which these irrigation impacts propagate downstream is dependent on the LA coupling processes as well as the irrigation method employed. This study has demonstrated that there are several key components necessary to effectively represent irrigation in coupled prediction models, including accurate land-cover classification, GVF, and an appropriate irrigation method and physics.
Similar impacts, in terms of soil moisture and fluxes (both offline and coupled) and feedbacks within the PBL, are expected should these generic irrigation approaches be applied to other LSMs. The extent to which an irrigated LSM spinup will impact the atmosphere in a coupled simulation is likely dependent on the details of LA coupling and model configuration used (physics options, resolution, etc.), but first-order impacts of introducing water to the land surface should be similar regardless of the LSM or coupled model used.
As the focus of this study is on the intercomparison of irrigation methods during the offline period and their cumulative impacts over the 5-yr spinup on the WRF initial condition, irrigation was not turned on in the coupled run. This allowed us to analyze the impact of the irrigated spinup alone on the physical processes while avoiding case study or time-dependent conclusions. The applicability of the coupled Drip results may be somewhat limited since the soil moisture remains largely unchanged, but this approach provided important results related to when and where, in terms of soil moisture and fluxes, the impacts of the Drip algorithm are manifested. Future work, especially that including longer-term and seasonal simulations, will use coupled irrigation.
Irrigation’s greatest impact on temperature and humidity occurs in regions that are both densely irrigated and downwind of other irrigated areas, because of the combination of direct and advected irrigation effects. In addition, the necessary conditions for cloud formation are most likely met if the LCL is low and the PBLH is high. Irrigation lowers both the PBLH and the LCL, resulting in competing effects on cloud formation. This study has found that in the absence of more dominant synoptic-scale forcing, irrigation results in both positive and negative feedbacks on the PBL depending on the time of day and the proximity to irrigation. Directly over irrigated areas, temperature drops, humidity rises, and the likelihood of surface-forced cloud development is increased in the morning, but the earlier PBL collapse breaks down the LoCo analyses late in the day. However, the advection of cooled and moistened air from irrigated areas reduces the LCL downwind but less directly impacts PBL growth, leading to greater chances of convective cloud development downwind of irrigation regardless of the time of day. These results may help explain the observational findings of Adegoke et al. (2007), who, using GOES infrared and visible images, detected peak convective cloud development that occurred 2 h earlier over cropland than over forested areas in Michigan on days featuring high pressure and light winds (<5 m s−1).
Even with a high-resolution simulation, evaluation of the irrigation methods with point observations proved to be difficult because of a number of factors, including the underrepresentation of irrigated areas in the USGS land-use classification data. Biophysical characteristics that determine transpiration amounts differ between crops, but the vegetation parameters in LIS–Noah do not account for different crop types. Thus, these land-use category differences not only complicated the point observations for the LSM evaluation by turning off irrigation and thus any differences between the methods at the Mead sites, but they also likely contributed to the underestimation of latent heat (through less ET) simulated by the model. Furthermore, this study used a climatological GVF, but in reality, phenology, and therefore ET, will vary based on the background climate conditions.
As the demand for food and fuel increases with a growing world population, the need to efficiently produce high crop yields will likely lead to further expansion of irrigated fields. The inclusion of irrigation physics then has the potential to improve forecasts, which will offer farmers a better tool to adapt to increasing crop demands. This study has shown that regional irrigation impacts are sensitive to time, space, and method and that irrigation cools and moistens the surface over and downwind of irrigated areas, ultimately resulting in both positive and negative feedbacks on the PBL. Future work will address these issues by using real-time GVF and a satellite-derived map of irrigated area (e.g., Ozdogan and Gutman 2008) and by addressing the interaction of irrigation with the assimilation of soil moisture in an LSM.
Acknowledgments
Much of this work was conducted as part of the NASA GSFC Intern Program and supported by the NASA Energy and Water Cycle Study (NEWS). Many thanks to the LIS team, especially Sujay Kumar, for providing feedback and support with LIS–WRF and LVT, and to Hiroko Beaudoing, Kristi Arsenault, and Eric Hunt for sharing their knowledge of irrigation and land-cover datasets. The MET analysis included data from the Research Data Archive (RDA; available via http://rda.ucar.edu/datasets/ds337.0), which is maintained by the Computational and Information Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR).
REFERENCES
Adegoke, J. O., Pielke R. A. Sr., Eastman J. , Mahmood R. , and Hubbard K. G. , 2003: Impact of irrigation on midsummer surface fluxes and temperature under dry synoptic conditions: A regional atmospheric model study of the U.S. high plains. Mon. Wea. Rev.,131, 556–564, doi:10.1175/1520-0493(2003)131<0556:IOIOMS>2.0.CO;2.
Adegoke, J. O., Pielke R. A. Sr., and Carleton A. M. , 2007: Observational and modeling studies of the impacts of agriculture-related land use change on planetary boundary layer processes in the central U.S. Agric. For. Meteor., 142, 203–215, doi:10.1016/j.agrformet.2006.07.013.
Betts, A. K., 1992: FIFE atmospheric boundary layer budget methods. J. Geophys. Res., 97, 18 523–18 531, doi:10.1029/91JD03172.
Bonfils, C., and Lobell D. , 2007: Empirical evidence for a recent slowdown in irrigation-induced cooling. Proc. Natl. Acad. Sci. USA, 104, 13 582–13 587, doi:10.1073/pnas.0700144104.
Boucher, O., Myhre G. , and Myhre A. , 2004: Direct human influence of irrigation on atmospheric water vapour and climate. Climate Dyn., 22, 597–603, doi:10.1007/s00382-004-0402-4.
Brakke, T. W., Verma S. B. , and Rosenberg N. J. , 1978: Local and regional components of sensible heat advection. J. Appl. Meteor., 17, 955–963, doi:10.1175/1520-0450(1978)017<0955:LARCOS>2.0.CO;2.
Chase, T. N., Piekle R. A. Sr., Kittel T. G. F. , Baron J. S. , and Stohlgren T. J. , 1999: Potential impacts on Colorado Rocky Mountain weather due to land use changes on the adjacent Great Plains. J. Geophys. Res., 104, 16 673–16 690, doi:10.1029/1999JD900118.
Chen, F., and Coauthors, 1996: Modeling of land surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res., 101, 7251–7268, doi:10.1029/95JD02165.
Chen, F., and Coauthors, 2007: Description and evaluation of the characteristics of the NCAR high-resolution land data assimilation system. J. Appl. Meteor. Climatol., 46, 694–713, doi:10.1175/JAM2463.1.
DeAngelis, A., Dominguez F. , Fan Y. , Robock A. , Kustu M. D. , and Robinson D. , 2010: Evidence of enhanced precipitation due to irrigation over the Great Plains of the United States. J. Geophys. Res., 115, D15115, doi:10.1029/2010JD013892.
Developmental Testbed Center, 2013: Model Evaluation Tools version 4.1 (METv4.1) user’s guide. Developmental Testbed Center, Boulder, CO, 226 pp. [Available online at www.dtcenter.org/met/users/docs/users_guide/MET_Users_Guide_v4.1.pdf.]
Douglas, E. M., Niyogi D. , Frolking S. , Yeluripati J. B. , Pielke R. A. Sr., Niyogi N. , Vörösmarty C. J. , and Mohanty U. C. , 2006: Changes in moisture and energy fluxes due to agricultural land use and irrigation in the Indian monsoon belt. Geophys. Res. Lett., 33, L14403, doi:10.1029/2006GL026550.
Evans, J. P., and Zaitchik B. F. , 2008: Modeling the large-scale water balance impact of different irrigation systems. Water Resour. Res., 44, W08448, doi:10.1029/2007WR006671.
Harding, K. J., and Snyder P. K. , 2012a: Modeling the atmospheric response to irrigation in the Great Plains. Part I: General impacts on precipitation and the energy budget. J. Hydrometeor., 13, 1667–1686, doi:10.1175/JHM-D-11-098.1.
Harding, K. J., and Snyder P. K. , 2012b: Modeling the atmospheric response to irrigation in the Great Plains. Part II: The precipitation of irrigated water and changes in precipitation recycling. J. Hydrometeor., 13, 1687–1703, doi:10.1175/JHM-D-11-099.1.
Huber, D., Mechem D. , and Brunsell N. , 2014: The effects of Great Plains irrigation on the surface energy balance, regional circulation, and precipitation. Climate, 2, 103–128, doi:10.3390/cli2020103.
Hutson, S. S., Barber N. L. , Kenny J. F. , Linsey K. S. , Lumia D. S. , and Maupin M. A. , 2004: Estimated use of water in the United States in 2000. USGS Circular 1268, 46 pp. [Available online at http://pubs.usgs.gov/circ/2004/circ1268/.]
Im, E., Marcella M. P. , and Eltahir E. A. B. , 2014: Impact of potential large-scale irrigation on the West African monsoon and its dependence on location of irrigated area. J. Climate, 27, 994–1009, doi:10.1175/JCLI-D-13-00290.1.
Irmak, A., Ratcliffe I. , Ranade P. , Hubbard K. , Singh R. K. , Kamble B. , and Kjaersgaard J. , 2011: Estimation of land surface evapotranspiration with a satellite remote sensing procedure. Great Plains Res., 21, 73–88.
Janjić, Z. I., 1994: The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927–945, doi:10.1175/1520-0493(1994)122<0927:TSMECM>2.0.CO;2.
Jiang, L., Ma E. , and Deng X. , 2014: Impacts of irrigation on the heat fluxes and near-surface temperature in an inland irrigation area of northern China. Energies, 7, 1300–1317, doi:10.3390/en7031300.
Koster, R. D., Dirmeyer P. A. , Hahmann A. N. , Ijpelaar R. , Tyahla L. , Cox P. , and Suarez M. J. , 2002: Comparing the degree of land–atmosphere interaction in four atmospheric general circulation models. J. Hydrometeor.,3, 363–375, doi:10.1175/1525-7541(2002)003<0363:CTDOLA>2.0.CO;2.
Kueppers, L. M., and Snyder M. A. , 2012: Influence of irrigated agriculture on diurnal surface energy and water fluxes, surface climate, and atmospheric circulation in California. Climate Dyn., 38, 1017–1029, doi:10.1007/s00382-011-1123-0.
Kueppers, L. M., Snyder M. A. , and Sloan L. C. , 2007: Irrigation cooling effect: Regional climate forcing by land-use change. Geophys. Res. Lett., 34, L03703, doi:10.1029/2006GL028679.
Kueppers, L. M., and Coauthors, 2008: Seasonal temperature responses to land-use change in the western United States. Global Planet. Change, 60, 250–264, doi:10.1016/j.gloplacha.2007.03.005.
Kumar, S. V., and Coauthors, 2006: Land information system : An interoperable framework for high resolution land surface modeling. Environ. Modell. Software, 21, 1402–1415, doi:10.1016/j.envsoft.2005.07.004.
Kumar, S. V., Peters-Lidard C. D. , Santanello J. , Harrison K. , Liu Y. , and Shaw M. , 2012: Land Surface Verification Toolkit (LVT)—A generalized framework for land surface model evaluation. Geosci. Model Dev., 5, 869–886, doi:10.5194/gmd-5-869-2012.
Lawrence, D. M., and Slingo J. M. , 2005: Weak land–atmosphere coupling strength in HadAM3: The role of soil moisture variability. J. Hydrometeor., 6, 670–680, doi:10.1175/JHM445.1.
Leng, G., Huang M. , Tang Q. , Gao H. , and Leung L. R. , 2014: Modeling the effects of groundwater-fed irrigation on terrestrial hydrology over the conterminous United States. J. Hydrometeor., 15, 957–972, doi:10.1175/JHM-D-13-049.1.
Lo, M.-H., and Famiglietti J. S. , 2013: Irrigation in California’s Central Valley strengthens the southwestern U.S. water cycle. Geophys. Res. Lett., 40, 301–306, doi:10.1002/grl.50108.
Lobell, D., Bala G. , Mirin A. , Phillips T. , Maxwell R. , and Rotman D. , 2009: Regional differences in the influence of irrigation on climate. J. Climate, 22, 2248–2255, doi:10.1175/2008JCLI2703.1.
Mahmood, R., Keeling T. , Foster S. A. , and Hubbard K. G. , 2013: Did irrigation impact 20th century air temperature in the high plains aquifer region? Appl. Geogr., 38, 11–21, doi:10.1016/j.apgeog.2012.11.002.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/BAMS-87-3-343.
Moore, N., and Rojstaczer S. , 2002: Irrigation’s influence on precipitation: Texas high plains, U.S.A. Geophys. Res. Lett., 29, doi:10.1029/2002GL014940.
NASS, 2009: 2007 census of agriculture: Farm and ranch irrigation survey (2008). Vol. 3, Special Studies, Part 1, Rep. AC-070SS-1, National Agricultural Statistics Service, USDA, 268 pp. [Available online at www.agcensus.usda.gov/Publications/2007/Online_Highlights/Farm_and_Ranch_Irrigation_Survey/.]
Oke, T. R., 1978: Boundary Layer Climates. Wiley, 372 pp.
Ozdogan, M., and Gutman G. , 2008: A new methodology to map irrigated areas using multi-temporal MODIS and ancillary data: An application example in the continental US. Remote Sens. Environ., 112, 3520–3537, doi:10.1016/j.rse.2008.04.010.
Ozdogan, M., Rodell M. , Beaudoing H. K. , and Toll D. L. , 2010: Simulating the effects of irrigation over the United States in a land surface model based on satellite-derived agricultural data. J. Hydrometeor., 11, 171–184, doi:10.1175/2009JHM1116.1.
Peters-Lidard, C. D., and Coauthors, 2015: Integrated modeling of aerosol, cloud, precipitation and land processes at satellite-resolved scales. Environ. Modell. Software, 67, 149–159, doi:10.1016/j.envsoft.2015.01.007.
Pokhrel, Y., Hanasaki N. , Koirala S. , Cho J. , Yeh P. J.-F. , Kim H. , Kanae S. , and Oki T. , 2012: Incorporating anthropogenic water regulation modules into a land surface model. J. Hydrometeor., 13, 255–269, doi:10.1175/JHM-D-11-013.1.
Qian, Y., Huang M. , Yang B. , and Berg L. K. , 2013: A modeling study of irrigation effects on surface fluxes and land–air–cloud interactions in the southern Great Plains. J. Hydrometeor., 14, 700–721, doi:10.1175/JHM-D-12-0134.1.
Santanello, J. A., Peters-Lidard C. D. , Kumar S. V. , Alonge C. , and Tao W.-K. , 2009: A modeling and observational framework for diagnosing local land–atmosphere coupling on diurnal time scales. J. Hydrometeor., 10, 577–599, doi:10.1175/2009JHM1066.1.
Santanello, J. A., Peters-Lidard C. D. , and Kumar S. V. , 2011a: Diagnosing the sensitivity of local land–atmosphere coupling via the soil moisture–boundary layer interaction. J. Hydrometeor., 12, 766–786, doi:10.1175/JHM-D-10-05014.1.
Santanello, J. A., and Coauthors, 2011b: Results from Local Land–Atmosphere Coupling (LoCo) Project. GEWEX News, Vol. 21, No. 4, International GEWEX Project Office, Silver Spring, MD, 7–9. [Available online at www.gewex.org/gewexnews/Nov2011.pdf.]
Santanello, J. A., Kumar S. V. , Peters-Lidard C. D. , Harrison K. , and Zhou S. , 2013a: Impact of land model calibration on coupled land–atmosphere prediction. J. Hydrometeor., 14, 1373–1400, doi:10.1175/JHM-D-12-0127.1.
Santanello, J. A., Peters-Lidard C. D. , Kennedy A. , and Kumar S. V. , 2013b: Diagnosing the nature of land–atmosphere coupling: A case study of dry/wet extremes in the U.S. southern Great Plains. J. Hydrometeor., 14, 3–24, doi:10.1175/JHM-D-12-023.1.
Segal, M., Pan Z. , Turner R. W. , and Takle E. S. , 1998: On the potential impact of irrigated areas in North America on summer rainfall caused by large-scale systems. J. Appl. Meteor., 37, 325–331, doi:10.1175/1520-0450-37.3.325.
Sivaraman, C., McFarlane S. , Chapman E. , Jensen M. , Toto T. , Liu S. , and Fischer M. , 2013: Planetary boundary layer (PBL) height value added product (VAP): Radisonde retrievals. U.S. DOE Rep. DOE/SC-ARM/TR-132, 36 pp. [Available online at www.arm.gov/publications/tech_reports/doe-sc-arm-tr-132.pdf?id=92.]
Skamarock, W. C., Klemp J. B. , Dudhia J. , Gill D. O. , Barker D. M. , Wang W. , and Powers J. G. , 2005: A description of the Advanced Research WRF version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp., doi:10.5065/D6DZ069T.
Sorooshian, S., Li J. , Hsu K. , and Gao X. , 2011: How significant is the impact of irrigation on the local hydroclimate in California’s Central Valley? Comparison of model results with ground and remote-sensing data. J. Geophys. Res., 116, D06102, doi:10.1029/2010JD014775.
Sorooshian, S., Klemp J. B. , Dudhia J. , and Gill D. O. , 2012: Influence of irrigation schemes used in regional climate models on evapotranspiration estimation: Results and comparative studies from California’s Central Valley agricultural regions. J. Geophys. Res., 117, D06107, doi:10.1029/2011JD016978.
Tang, Q., Peterson S. , Cuenca R. H. , Hagimoto Y. , and Lettenmaier D. P. , 2009: Satellite-based near-real-time estimation of irrigated crop water consumption. J. Geophys. Res., 114, D05114, doi:10.1029/2008JD010854.
Tuinenburg, O. A., Hutjes R. W. A. , Stacke T. , Wiltshire A. , and Lucas-Picher P. , 2014: Effects of irrigation in India on the atmospheric water budget. J. Hydrometeor., 15, 1028–1050, doi:10.1175/JHM-D-13-078.1.
Vahmani, P., and Hogue T. S. , 2014: Incorporating an urban irrigation module into the Noah land surface model coupled with an urban canopy model. J. Hydrometeor., 15, 1440–1456, doi:10.1175/JHM-D-13-0121.1.
Wei, J., Dirmeyer P. A. , Wisser D. , Bosilovich M. G. , and Mocko D. M. , 2013: Where does the irrigation water go? An estimate of the contribution of irrigation to precipitation using MERRA. J. Hydrometeor., 14, 275–289, doi:10.1175/JHM-D-12-079.1.
Yilmaz, M. T., Anderson M. C. , Zaitchik B. , Hain C. R. , Crow W. T. , Ozdogan M. , Chun J. A. , and Evans J. , 2014: Comparison of prognostic and diagnostic surface flux modeling approaches over the Nile River basin. Water Resour. Res., 50, 386–408, doi:10.1002/2013WR014194.
Zaitchik, B. F., Evans J. , and Smith R. B. , 2005: MODIS-derived boundary conditions for a mesoscale climate model: Application to irrigated agriculture in the Euphrates basin. Mon. Wea. Rev., 133, 1727–1743, doi:10.1175/MWR2947.1.
