Effect of ENSO-based upstream water withdrawals for irrigation on downstream water withdrawals

In the Southeast US, El Niño Southern Oscillation (ENSO), climate variability phenomena affect the quantity of water that is available for irrigation. The goals of this study were to determine the effect of upstream surface water withdrawals for irrigation on the quantity of water available for irrigation in downstream areas as a function of the ENSO phase and quantify the watershed area that can be irrigated using water withdrawn from streams in an ecologically sustainable manner. The study was conducted in the Swan Creek watershed (97 km) located in Limestone County, Alabama, USA. The soil and water assessment tool (SWAT) model was used to simulate stream flows and develop water withdrawal prescriptions. Results indicated that when simultaneous water withdrawals were made at the outlet of each subwatershed throughout the year, on average water withdrawals were sufficient to irrigate 4–16% of the area upstream of withdrawal point depending on stream order. On making sustainable withdrawals at the outlets of all subwatersheds and at the watershed outlet throughout the year, approximately 40% of the watershed area could be irrigated. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2020.156 om http://iwaponline.com/hr/article-pdf/51/4/602/730616/nh0510602.pdf 2020 Laljeet Sangha Jasmeet Lamba (corresponding author) Hemendra Kumar Biosystems Engineering Department, Auburn University, Auburn, AL 36849, USA E-mail: jsl0005@auburn.edu


INTRODUCTION
The world population is estimated to grow to 8.3 billion in 2030 and 9.3 billion in 2050 with nearly 67 million people being added to the world per year (FAO ). To meet the needs of the projected population, crop production should increase by nearly 50% in the next 50 years to sustain our present per capita supply, assuming the productivity of present farmland remains the same (Jury & Vaux ). Irrigation can help producers to sustain and increase crop production (Veldkamp et al. ), especially in areas that receive a limited amount of rainfall during the crop growing season.
Water withdrawn from streams is one of the major sources of irrigation in southeastern US. Water is typically withdrawn from the streams and stored in on-farm ponds for irrigation. However, it is important that water withdrawals from streams for irrigation should be done in an ecologically sustainable manner. Excessive water withdrawals from streams disturbs in-stream biota by reducing functioning habitat (Scatena & Johnson ), blocking the entrance to habitat, and causing direct and indirect mortality (Benstead et al. ). Therefore, for effective water management, efficient and planned water withdrawals are required for irrigation. Hydrological models capable of simulating watershed level hydrological processes on a long-term basis and at the subwatershed level can help evaluate the effects of management decisions (e.g., water withdrawals from streams) on water resources (e.g., levels of streamflows) (Douglas et al. ; Gassman et al. ).
Irrigation water demand is expected to rise in the future due to anticipated variations in the rainfall regime caused by climate variability (Díaz et al. ). Climate variability in the southeastern US is governed by El Niño Southern Oscillation (ENSO) phenomenon. ENSO refers to the year-to-year variation in surface air pressure, sea surface temperatures, convective rainfall, and atmospheric circulation that appears over the equatorial Pacific Ocean (Philander ). Thus, the volume of water available for withdrawals from streams is also a function of the ENSO phase. Ropelewski & Halpert () found that the influence of ENSO on precipitation in the southeastern US is spatially less consistent.
This was also confirmed in a study by Sharda et al. (), in which opposite correlations were found between ENSO and precipitation, and ENSO and streamflow patterns between northern and southern Alabama (AL). Overall, previous studies have shown that the impact of ENSO on precipitation and streamflows cannot be generalized over larger areas.
The instant effect of water withdrawal from streams is a drop in stream water levels in the downstream areas, which differs within a watershed (Henderson ; Lai et al. ).
Water withdrawn from streams by farmers for irrigation in upstream areas can reduce the volume of water available for withdrawal in downstream areas. This will not only be harmful to downstream biota, but could also turn out to be an economic disaster for farmers who would have access to a limited amount of water available for withdrawal to irrigate crops. Therefore, it is important to consider how streamflow withdrawal in upstream areas of watershed impact streamflow in downstream areas. Mondal et al.
() conducted a study in a forested watershed in south AL and quantified the area within a watershed that can be irrigated using water withdrawn from streams. However, Mondal et al. () assumed that the withdrawals are made only at the outlet of a particular stream order at a time. Typically, in agricultural watersheds, water withdrawals are made simultaneously at the outlets of various subwatersheds at a time for irrigation (e.g., multiple farmers withdrawing water from streams for irrigation). To our knowledge, no study has evaluated the effect of upstream surface water withdrawals for irrigation on the quantity of water available for irrigation in downstream areas as a function of the ENSO phase in agricultural watersheds. This study aims to quantify: (a) the effects of ENSO on precipitation and streamflow during crop growing and noncrop growing seasons, (b) the impact of upstream water withdrawals on the downstream water withdrawals as a function of the ENSO phase, and (c) the watershed area that can be irrigated using water withdrawn from streams in an ecologically sustainable manner. The research results from this study will provide a valuable dataset for conservation planners that can be used to plan water withdrawals from streams for irrigation without disturbing the ecological integrity of streams.

Study area
The study watershed was Swan Creek watershed (97 km 2 ), which is a part of the larger Tennessee river basin. The water-

Soil and water assessment tool model
The soil and water assessment tool (SWAT) model has proven to be a useful tool for evaluating water resource problems for a wide range of watershed scales and environmental conditions across the globe (Francesconi et al. ). The model is physically based, computationally efficient, and can simulate hydrological processes over long periods. Hydrology, weather, soil properties, plant growth, nutrients, and land management are the major

Data input
Topographical data were obtained using a 10-m digital elevation model (DEM), which was obtained from the USDA-NRCS National Geospatial Data Gateway (https:// datagateway.nrcs.usda.gov/). Soil data were obtained from the Soil Survey and Geographic database (NRCS ).
Planting date, tillage methods, timing and rate of nutrient and pesticide applications, and harvest timing were obtained  (Table 1) identified from the previous scientific studies were changed to achieve maximum agreement between observed and simulated flows. Based on the availability of the observed streamflow data, the model was    Water was withdrawn from all the streams in a chronological manner in the watershed (i.e., water was withdrawn from first-order streams, then second-order followed by third-order streams). The water was withdrawn using the criteria explained previously in the manuscript. The certain volume of water was withdrawn from the outlet of each subwatershed while maintaining the minimum levels in the streams required to sustain the in-stream ecology. One of the limitations of the SWAT model is that it does not allow the amount of water withdrawn from the stream to vary at a daily time-step. Therefore, water withdrawal analysis was done outside the model using variables from the reach (.rch) and subwatershed (.sub) output files of the model.

RESULTS AND DISCUSSION
Calibration and validation of the SWAT model which would assert the need for irrigation water withdrawal management practices to obtain the optimum amount of water for irrigation in an ecologically sustainable manner.

Relationship between ENSO and streamflow
Similar to precipitation trends, in the noncrop growing season, the streamflows at the watershed outlet in the La Niña phase were greater than the El Niño phase Standard error bars are at 5% of error.
( Figure 4(a)). This trend was opposite that observed by   Niña phases during the crop growing season ( Figure 5(b)).
However, it should also be noted that even though more water is available for withdrawal during the El Niño phase in the crop growing season, the volume of water available for withdrawal during the noncrop growing season in the El Niño phase was almost three times the volume of water available for withdrawal during the crop growing season.
Thus, the results of this study show that the winter

First-order streams
The outlets of first-order stream subwatersheds such as 1, 2, 3,6,8,9,10,11,14, and 18 ( Figure 6) were the foremost points of water withdrawals in the watershed. When the withdrawals from the stream were made throughout the year, on average 16% of the area upstream of withdrawal point could be irrigated (Table 3). It was observed that on average, the water withdrawn in an ecologically sustainable manner throughout the year at the outlet of first-order stream subwatershed was sufficient to irrigate 109 × 10 4 m 2 (269 acres) (Table 3). If the water withdrawals were only made in the noncrop growing season, on average 10% of the area (i.e., 64 × 10 4 m 2 ) upstream of withdrawal point could be irrigated by the water withdrawn (Table 3

Second-order streams
The outlets of second-order stream subwatersheds such as 4, 5, 7, 12, 13, 15, 16, and 17 ( Figure 6) were succeeding withdrawal points after the withdrawals have been made at the outlet of first-order stream subwatersheds. Due to the water withdrawals made at the outlets of first-order stream subwatersheds, the amount of water available for withdrawal at the outlets of second-order stream subwatersheds was less than natural flows (i.e., when water was not withdrawn from first-order streams). When the sustainable water withdrawals were made throughout the year from the outlets of all the secondorder stream subwatersheds, on average 8% of the area (i.e., 298 × 10 4 m 2 (736 acres)) upstream of second-order stream subwatershed could be irrigated by the stream water withdrawals (Table 3). It should be noted that on average 109 × 10 4 m 2 (269 acres) of the upland area has already been irrigated by the water withdrawn at the outlet of first-order stream subwatersheds. Therefore, when water is withdrawn throughout the year, the total area irrigated by the water withdrawals made at the outlet of firstand second-order stream subwatersheds was found to be 3,481 × 10 4 m 2 (8,601 acres), which was approximately 35% of the total watershed area. On average on making the simultaneous withdrawals at each firstand second-order streams, 1,254 × 10 4 m 2 (3,098 acres), of the area upstream of the withdrawal point, could be irrigated.
This was approximately 13% of the area upstream of the withdrawal points. During the noncrop growing season, the total area irrigated by firstand second-order stream was 2,024 × 10 4 m 2 (5,000 acres), which was 20% of the watershed area. to less precipitation in our study watershed.

Third-order stream
At the watershed outlet (i.e., third-order stream), when water was withdrawn throughout the year, the stream water withdrawal was enough to irrigate an average 4% of the area upstream of withdrawal point (Table 3) Water withdrawal only at the outlet of second-order streams (Scenario 3) Water was withdrawn at the outlet of all second-order stream subwatersheds (stream 4, 5, 7, 12, 13, 15, 16, 17, 19). Flows in the first-order streams were left undisturbed which resulted in higher flows in second-order streams.
Therefore, on average, when the water withdrawals were made throughout the year, the second-order streams could irrigate 11% of the area upstream of the withdrawal point compared with 8% in scenario 2 (when the water was withdrawn from the outlet of first-order streams). On making the withdrawals throughout the year, at the watershed outlet, on average, the volume of water available for withdrawal was greater than the second scenario ( Figure 7). When no water was withdrawn from first-order streams, on average, 6% of the area upstream of third-order stream subwatershed could be irrigated by the water withdrawn throughout the year. Similar trends were observed for water withdrawals made in crop growing and noncrop growing seasons ( Figure 7). Compared with Scenario 2, when water was withdrawn simultaneously at the outlet of firstand second-order stream subwatersheds, the percentage of the total area that could be irrigated upstream of second-order stream subwa- Therefore, water withdrawal strategies should be planned according to the ENSO phase which can be predicted in advance. This will allow farmers to plan and withdraw water from streams sustainably for irrigation.

SUMMARY AND CONCLUSIONS
The results of the study indicate that precipitation and