Rain-fed to irrigation-fed transition of agriculture exacerbates meteorological drought in cropped regions but moderates elsewhere

In recent decades, irrigated agriculture has expanded dramatically over the Southeastern United States (SEUS). The trend is more likely to continue in future given the need to further improve crop productivity and its resilience against droughts, however, the impact of these SEUS land cover changes remains unknown. This study investigates how and to what extent rain-fed to irrigation-fed (RFtoIF) transition in the SEUS region modulates precipitation spatially and temporally under a severe drought meteorological condition. In this study, we perform three Weather Research Forecasting model simulations with varying degrees of irrigated crop areas with meteorological boundary conditions of a record-breaking 2007 drought in the SEUS region. Results show that the SEUS irrigation expansion reduces both the convective triggering potential and low-level humidity index through land-atmospheric interaction. This is accompanied by reduction in the height of atmospheric boundary layer (ABL)-lifting condensation level crossing and increase in the convective available potential energy. These modulations within the ABL provide a favorable condition for strong deep convection during the drought period. However, the impact on precipitation is heterogeneous, with crop areas undergoing RFtoIF transition experiencing an overall reduction in precipitation while other landcovers experiencing an increase. The reduction in precipitation over RFtoIF transitioned croplands is in part due to moisture redistribution aided by generation of an anomalous high-pressure system. The results highlight the complexity of response of precipitation to irrigation expansion in the SEUS, and underscore the need to perform spatially-explicit analysis for mitigating risks to water resources and food security.


Introduction
Irrigation continues to expand through the cropped regions all over world to meet the growing demand of food and fiber (Pervez andBrown 2010, Siebert et al 2015).Irrigation-fed agriculture already provides for roughly half of the total value of U.S. crop production on 28% of cropland (USDA 2019, Hrozencik andAillery 2021).Irrigated farmland in the US climbed to 58 million acres in 2017, a 4% increase from 2012 (Nass 2017).In recent decades, significant irrigation expansion has occurred over the humid southeastern US states.For example, irrigated acreage has increased by +46% in Georgia and +212% in Tennessee over 1997-2012(Walton 2019)).
Irrigation expansion can alter a range of coupled environmental states and fluxes.Groundwater pumping for irrigation changes the water table level and base flow (Condon and Maxwell 2019).Its impact on local meteorology has been often studied through the lens of land-atmosphere interactions.Irrigation enhances soil moisture and the water vapor content of the near-surface, but decreases the surface and nearsurface temperature (Sacks et al 2009, Ozdogan et al 2010, Harding and Snyder 2012, Qian et al 2013, Wei et al 2013, Lu et al 2017).Irrigation expansion modulates the land-atmosphere interactions and other atmospheric processes, resulting in changes in precipitation, however the eventual impacts are oftentimes varied depending on the local hydroclimatic and physiographic conditions (Harding and Snyder 2012).For example, previous studies have reported both deficit or surplus in precipitation due to irrigation (Saeed et al 2009, Deangelis et al 2010, Puma and Cook 2010, Harding and Snyder 2012, Wei et al 2013, Huber et al 2014, Pei et al 2016).
Given the ongoing rain-fed to irrigation-fed (RFtoIF) transition of croplands in Southeastern United States (SEUS), there is a crucial need to assess how such a RFtoIF transition affects regional precipitation patterns.Specifically, the study aims to better understand the role of irrigation on the landatmosphere coupling and the modulation of vertical mixing processes and associated large-scale circulations at the sub-seasonal scale during the summer of 2007 when SEUS experienced an extreme meteorological drought.To this end, three Weather Research Forecasting (WRF) model runs for summer (June-August) of 2007 were performed.The WRF model runs consist of a control run with no irrigation on croplands and the two experiment runs with different irrigation expansion perturbations.Through these runs, we answer two scientific questions: (1) how may the RFtoIF transition modulate the landatmosphere interaction during 2007-like droughts in the SEUS, and (2) will RFtoIF transition impact the precipitation magnitude and pattern over the region?By answering these questions, this study will advance the understanding of the atmospheric response to irrigation expansion.

Study area
The study was conducted for the SEUS states encompassing all of Alabama, Arkansas, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee (figure 1(a)).Some areas of neighboring states were also included.We specifically focus our analysis on the Deep South region that includes Alabama, Georgia, and Mississippi.Notably, despite being a water ample region, the SEUS has experienced severe droughts in recent decades.Over 2006-2008, the SEUS states experienced a severe drought with a peak in 2007, primarily driven by precipitation deficit (Kam et al 2014).This drought caused economic losses of over $1 billion and resulted in lack of available water resources along rivers and in lakes, triggering inter-basin water import to the region for the first time in 100 years (Manuel 2008, Campana et al 2012).Growing population (PNREAP 2023, USAFacts 2023) and expansion of irrigationfed agriculture make the regional communities more vulnerable to droughts.There is a need to refine our understanding of whether the anticipated water scarcity risks from meteorological droughts in SEUS will be elevated or suppressed by the indirect impacts of RFtoIF transition on precipitation.

Experiment design
In this study, Weather Research and Forecasting model version 4 (WRF v4.0) was implemented in the SEUS for the three summer months (June-August) of 2007.The concerned period represents the driest summer during the severest SEUS drought over the last 50 years.The mother and inner model domains discretized the SEUS (domain-01) at 15 km resolution, while the nested Deep South (domain-02) was discretized at 3 km resolution, respectively (figure 1(a)).The inner model domain considers a much-resolved convection-permitting formulation, while the mother model domain uses cumulus physics for computational efficiency (see table S1 for the details of the model configuration).
We used the North American Regional Reanalysis (NARR) dataset (Mesinger et al 2006) for initial and lateral boundary conditions of our WRF runs.Default land category data is from the National Land Cover Database (NLCD) of 2006 (Homer et al 2020).All cultivated crops and pasture/hay NLCD land categories were defined as croplands for the ensuing analysis.The croplands included both rain-fed and irrigated lands.The irrigation-fed region was selected based on the Moderate Resolution Imaging Spectroradiometer Irrigated Agriculture Datasets for the Conterminous United States (MIrAD-US) Version 4 for 2007 at the 1 km spatial resolution (Shrestha et al 2021).We used the Pu-Xleim land surface model (Gilliam and Pleim 2010) that includes shallow (surface to 1 cm) and root-zone (1-99 cm) soil layers.
A spin-up simulation was performed for soil moisture stabilization over May 2007 to generate a more realistic initial soil moisture conditions.To simulate the effects of irrigation, we forced the soil moisture in the 2 soil layers of all irrigated land to be fully saturated during the entire study period.The control (CTL) run was the default WRF simulation where it was assumed that all croplands are rainfed.For the Experiment 1 (EXP1) run, soil moisture was set to be fully saturated over the irrigation-fed regions.For Experiment 2 (EXP2) run, all croplands, including rain-fed, were assumed to be irrigation-fed.Again, soil moisture was set to full saturation over all irrigation-fed regions.The difference in soil moisture due to RFtoIF transition is well apparent in figures 1(b) and (c).

Analysis of land-atmosphere coupling
The land-atmosphere coupling was diagnosed via the convective triggering potential (CTP) and lowlevel humidity index (HI LOW ) framework (Findell and Eltahir 2003a, Ferguson and Wood 2011, Jach et al 2020, 2022).HI LOW ( • C) is calculated by the sum of the dewpoint depressions at 50 and 150 hPa above ground level (AGL) at 6 local sidereal time: where T and T d are temperature ( • C) and dewpoint temperature ( • C), respectively.Subscript of AGL-p indicates the pressure level p AGL. CTP (J kg −1 ) is obtained by integrating vertical profile between the moist temperature of air parcel, T m (K) and the environmental temperature, T e (K) from 100 to 300 hPa AGL at the same time: where g 0 is the gravitational acceleration (9.81 m s −2 ).The height of z AGL−100 and z AGL−300 are located near 1 km and 3 km, respectively.The CTP-HI LOW framework classifies landatmosphere feedback into four categories: (1) dry soil advantage, (2), wet soil advantage, (3) transition and (4) atmospherically controlled.The dry soil advantage regime designates thermal triggering of precipitation wherein high sensible heat flux leads to boundary layer growth and upward mixing of moist air to heights where condensation and formation of rainfall can occur (Dirmeyer et al 2014).The wet soil advantage regime specifies hydrologic triggering of precipitation wherein high soil moisture wets the boundary layer thus increasing the predisposition of condensation as the moist air rises.In the transition regime, convection may be triggered in either wet or dry soils.Notably, this configuration seldomly leads to precipitation.The atmospherically controlled regime inhibits the contribution of land towards triggering deep convection.An atmospherically controlled regime may also be termed as 'too dry for rain' and 'too stable for rain' .

Impact of irrigation on land atmosphere interaction
For the summer (June-August) of 2007, the regional averages of simulated monthly 2 m temperature and relative humidity by the three control and experiment WRF runs were evaluated against those from the NARR dataset (supplementary figure S1).The EXP1 run with recent land cover and irrigation area status shows hot and dry bias compared with the NARR data.For example, simulated 2 m temperature in the EXP1 run was slightly overestimated.The EXP1 run underestimated both precipitation (by −0.87 mm d −1 or −25.32%) and 2 m RH (by −3.15%).The negative bias of precipitation in the WRF simulation was in line with similar results from previous studies using the WRF model (Prein Estimates of CTP and HI LOW showed that the averages of CTP-HI LOW in the Deep South were under the dry soil advantage regime during the summer of 2007 (figure 2).It is worth noting that the SEUS region is, however, usually defined as the wet soil advantage regime (Findell and Eltahir 2003b).Furthermore, RFtoIF transition led to a more humid near-surface and more stable atmospheric boundary layer (ABL).For example, the regional averages of HI LOW in EXPs were lower than that in CTL due to a combination of modulation of the surface energy budget (e.g.increased latent heat and decreased sensible heat fluxes).Humid near surface condition also interacted with ABL by making the humid air temperature profile become closer to wet adiabatic.This ABL condition decreased the CTP in EXPs, contributing to a more stable atmospheric condition.With irrigation expansion, the days under the 'atmospherically controlled: too dry for rain' regime (over 15 • C of HI LOW ) decreased from 33 d in the CTL run to 29 and 22 d in the EXP1 and EXP2 runs, respectively.The days under the 'wet soil advantage' regime, however, increased from 31 d in the CTL run to 35 and 49 d in EXP1 and EXP2 runs, respectively.

Impact of irrigation on cloud formation and precipitation
Results show that irrigation expansion modulated the vertical distribution of clouds and precipitation via land-atmosphere interactions.Clouds can develop at the height where/when ABL and the lifting condensation level (LCL) intersect.This crossing determines the initiation of cloud formation and convection triggering (Gentine et al 2013, Yin et al 2015).In the EXP runs, irrigation increased the soil moisture, and consequently the land-atmosphere coupling decreased the height of the ABL-LCL crossing (figure 3(a)).In contrast, the CTL run (drier rain-fed cropland) showed relatively larger sensible heat flux into ABL, consequently deepening the ABL height, which in turn caused a higher ABL-LCL crossing.The difference in height of ABL-LCL crossing visà-vis the wetness is consistent with observational (Phillips and Klein 2014) and 1D ideal model (Yin et al 2015) studies.Our WRF runs also showed an earlier timing of ABL-LCL crossing for dryer land case The impact on precipitation is heterogeneous and does not follow the spatial distribution of 2 m temperature, 2 m relative humidity, and soil moisture (supplementary figure S2).For example, the transition from RFtoIF increased soil moisture over the western Mississippi and Tennessee and the southwestern Georgia.However, the spatial pattern of precipitation is much more heterogeneous (figures 4(a) and (b)).Differences between vertically-integrated moisture flux (figures 4(c) and (d)) showed that irrigation transition induced an anomalous high pressure system.The high pressure system became stronger from EXP1 to EXP2 because of enhanced cooling of the land and subsequently the near-surface atmosphere.This high pressure system played a role in altering the spatial distribution of atmospheric moisture, with increased precipitation over parts of Arkansas, Louisiana, and Florida in EXP2.In contrast, moisture fluxes over western South Carolina and croplands of southwestern Georgia impeded the moisture coming in from the eastern shores of the SEUS, resulting in suppressed precipitation.Croplands along the Mississippi River also experienced precipitation suppression due to being near the edge of the anomalous high system.This is of significance as the region is already experiencing groundwater depletion to support irrigated agriculture (Reba et al 2017).Further RFtoIF transition in the SEUS may put extra pressure on the groundwater aquifer along the lower Mississippi Basin due to further reduction in precipitation.Overall, the amount of precipitation is found to decrease over croplands but increase over noncropped areas.The disparity in precipitation between EXP2 and CTL is statistically significant at 5% level based on the two-tailed Student's t-test.In terms of the overall atmospheric moisture balance, both crop and non-crop areas experienced a decrease in atmospheric moisture (supplementary figure S3(c)).This is true even though evapotranspiration minus precipitation increased with RFtoIF extent, partially due to increase in evapotranspiration from irrigated cropped areas and also an because of increase in net radiation in non-cropped lands caused by suppressed cloudiness and precipitation conditions resulting from the formation of the high-pressure system (supplementary figure S4).The overall reduction in atmospheric moisture is mainly attributable to a greater divergence of moisture out of the region with increasing extent of RFtoIF transition (supplementary figure S3).This is also consistent with the formation of anomalous high-pressure system in EXP1 and EXP2 (see figure 4).In terms of the total regional precipitation during the summer season, relatively small RFtoIF transition results in an overall enhanced precipitation in the SEUS region, but full RFtoIF transition of all croplands negates the overall increase (supplementary figure S5).

Conclusions and synthesis
Southeastern US has been experiencing RFtoIF transition of agriculture at a faster pace than ever, spurred by a range of socioeconomic impacts including record corn and soybean prices.An amendment in the 2014 US farm bill has also facilitated this expansion by allowing non-western states to apply for federal irrigation grants.The 2007 SEUS drought that caused billions of crop losses, further underscored the need to adjust the level of perceived risk of local stakeholders and policy makers to droughts through irrigation expansion.Through RFtoIF transition, there is a potential for SEUS farmers to increase crop productivity thus yielding economic benefits.Furthermore, the RFtoIF transition would help farmers and insurers manage risks against erratic rainfall, particularly during the dry spells.However, given that RFtoIF transitions and associated modulation of land-atmosphere interaction is likely to affect land-atmosphere coupling, moisture recycling, atmospheric circulation, and precipitation distribution across the scales, assessment of advantages of RFtoIF transition on food and water security, especially during droughts, remain uncertain.This study assessed the impact of RFtoIF transition on modulating precipitation during droughts.
Results showed that irrigation expansion decreased near-surface temperature and increasing nea-surface humidity, which in turn led to reduction of HI LOW and CTP.In other words, ABL became wetter and more stable.This implies that irrigation weakens dry land-atmosphere coupling in the SEUS region during severe droughts.Despite the smaller potential for convective motion within the ABL, the potential of deep convection was enhanced.However, increased potential of deep convection did not increase precipitation everywhere.The east side of Deep South (e.g.southwestern Georgia) experienced precipitation deficits due to the blocking of moisture transport via anti-cyclone air motions over anomalous near-surface high pressure regions.
Our results were based on a set of WRF simulations.As is the case with most model results, the reported magnitude of the changes has inherent uncertainties.However, the physical consistency of the results between the CTL and EXP runs elicits confidence in their validity.Further confidence in the results could be obtained through the usage of large ensemble members of control and experiment runs, although, the computational demand for such simulations remain prohibitive and will need to be addressed in future studies.In the EXP runs, this study assumed that irrigation was applied in a manner that ensured persistent saturated soil conditions thus minimizing crop stress.Future studies may focus on further elucidating the impacts of different intensities and schedules of irrigation on precipitation alterations.
Despite these limitations, the study demonstrates the impacts of RFtoIF transition in SEUS.Overall, our results imply that, spatially, irrigation expansion may have divergent impacts depending on the area under consideration, with some locations experiencing an increase in precipitation but others (especially cropped regions) experiencing a decrease.The increase is largely due to enhanced source of moisture from irrigation and development of favorable conditions in the lower atmosphere for precipitation occurrence.The decrease is caused by blocking of the moisture flux in the affected regions due to generation of a high-pressure system.The study also highlights a non-monotonic influence on amount of precipitation with RFtoIF transition.In terms of the total precipitation during the summer, relatively small RFtoIF transition (i.e.equivalent to the current irrigated area fraction) results in an overall enhanced precipitation in the region, which can be a welcome change for water users and managers, particularly during a severe drought.However, a full RFtoIF transition, wherein all croplands in the region are transitioned to irrigated land, negates the overall increase due to blocking of incoming moisture flux in the affected areas.
The study shows that widespread irrigation expansion may even exacerbate precipitation deficit locally.Given these heterogenous impacts on precipitation, it is important to plan water and food security risk mitigation measures that account for spatially-explicit impacts of regional RFtoIF transitions.

Figure 1 .
Figure 1.(a) Rain-fed (blue) and irrigation-fed (green) regions in WRF simulations.Frame boundary (grey box) and the red box indicate domain-01 and domain-02, respectively.The 3 month-averaged (from June to August) soil moisture (mm) for (b) rain-fed (control run), and (c) all irrigation-fed (Experiment2) simulations.Spatially-averaged diurnal cycle of (d) latent heat flux (W m −2 ) and (e) sensible heat flux (W m −2 ) in domain-02 for Control (CTL), Experiment1 (EXP1), and Experiment2 (EXP2) are shown using grey dots, and green and blue lines, respectively.CTL, EXP1, and EXP2 are WRF simulation configurations where all croplands are considered rain-fed, rain-fed and irrigated-fed distribution of croplands is as per the data in the year 2007, and all croplands are considered irrigation-fed, respectively.LST is local sidereal time.

Figure 2 .
Figure 2. Scatter plot of the spatial average of convective trigger potential (CTP; J kg −1 ) and low-level humidity index (HILOW; • C), over the Deep South (domain-02) for (a) Control run (CTL; grey), (b) Experiment1 (EXP1; green), and (c) Experiment2 (EXP2; blue).Configuration summary of CTL, EXP1, and EXP2 WRF is presented in figure 1 caption.Each open black dot is a daily average from the simulation.Error bars (in (a)-(c)) denote the one standard deviation of the daily average.Red, blue, and grey rectangles in the background denote regimes of dry soil advantage (Dry), wet soil advantage (Wet), and transition (Trans).White background indicates atmospherically controlled regimes.

Figure 3 .
Figure 3. (a) Time evolution of atmospheric boundary layer (ABL; km) height and lifting condensation level (LCL; km) for control (CTL), experiment1 (EXP1), and experiment2 (EXP2) WRF simulations.Configuration summary of CTL, EXP1, and EXP2 is presented in figure 1 caption.ABLs from CTL and EXPs are x marks and dash-dot lines, respectively, while LCLs are shown as filled dots and solid lines.(b) Maximum convective available potential energy (CAPEMAX; J kg −1 ) from CTL, EXP1, and EXP2.Error bars indicate one standard error of spatiotemporal average from runs.(c) Vertical profiles of relative humidity.Grey, green, and blue correspond to CTL, EXP1, and EXP2 runs, respectively.LST is local sidereal time.

Figure 4 .
Figure 4. Spatial distribution of average differences for (a), (b) precipitation (mm day −1 ) and (c), (d) vertically-integrated moisture flux (Q flux, vector; kg m s −1 ) for EXP1-CTL (left) and EXP2-CTL (right).Configuration summary of CTL, EXP1, and EXP2 is presented in figure 1 caption.The dotted area in (a) and (b) indicates the statistical significance at 5% level from the two-tailed Student's t-test.'H' in (c) and (d) indicate the high-pressure anomalous system.Abbreviations indicate the names of states: Arkansas (AR), Tennessee (TN), North Carolina (NC), South Carolina (SC), Mississippi (MS), Alabama (AL), Georgia (GA), Louisiana (LA), and Florida (FL).(e) The amount of accumulated precipitation (mm day −1 ) under total, crops, and non-crops fraction area for CTL (grey bars), EXP1 (green bars), and EXP2 (blue bars).Crop area includes both rain-fed and irrigation-fed.The rest of the area is listed as non-crops.The total area equals sum of crops and non-crops.

Figure S4 .
Figure S4.Average net radiation on the surface (W m -2 ) in different regions.
Beltrán-przekurat A, Niyogi D, Pielke Sr R and Vörösmarty C 2009 The impact of agricultural intensification and irrigation on land-atmosphere interactions and Indian monsoon precipitation-a mesoscale modeling perspective Glob.Planet.Change 67 117-28 Ferguson C R and Wood E F 2011 Observed land-atmosphere coupling from satellite remote sensing and reanalysis J. Hydrometeorol.12 1221-54 Findell K L and Eltahir E A 2003a Atmospheric controls on soil moisture-boundary layer interactions.Part I: framework development J. Hydrometeorol. 4 552-69