The Sunflower River Flow Augmentation Project model

From GeoMod

Project Goals

The aim of this project is to determine the effects of the Sunflower River Flow Augmentation Project on adjacent water levels.

Why augment the Sunflower River

Decreasing flow in the river

Lowest recorded flows from 1950 to 1990 on the Sunflower River at Sunflower, MS. Provided by YMD.

There have been significant declines in the flow in the Sunflower River over the last 25 years.

Lowest flows occur in the late summer and early fall when groundwater pumping for irrigation is at its maximum.

USGS data: USGS 07288500 BIG SUNFLOWER RIVER AT SUNFLOWER, MS (Note: data from 1980-2001 missing)

Falling water table levels

Groundwater level changes in the alluvial aquifer from October 1995 to October 2005. Provided by YMD.

Irrigation water comes primarily from groundwater pumping.

• 98% of water pumped is for agricultural irrigation (Arthur, 2001)
• Approx. 17,000 agricultural irrigation wells screened in the alluvial aquifer.

Induced stream discharge into aquifer

Diagram showing the effects of ground water pumping on base flow to streams. Based on Winter et al., (1998).

At peak pumping times water table levels drop beneath the river changing this gaining stream to a losing stream.

Illustration of the Sunflower River’s interaction with the alluvial aquifer. Image from USGS Groundwater Atlas of the United States (Renken, 1998).

Disappearing streams in Kansas

Loss of perennial streams in Kansas between 1961 and 1994 after Sophocleous (2000).

Because the surface and ground water systems are tightly integrated, stream loss is a big problem everywhere there is substantial use of the water table aquifer.

What is the Sunflower River Flow Augmentation Project

Sunflower River Flow Augmentation Project location in northwestern Coahoma County, MS. Location in Google Maps

The Sunflower River Flow Augmentation Project is a well field that pumps groundwater from the alluvial aquifer and uses it to supplement flows in the Sunflower River.

• Located approximately 34.380N, -90.598W (Google Map)
• Between the Mississippi River and the Yazoo River.

Groundwater, extracted from the alluvial aquifer is discharged into a canal which feeds into Long Lake. Water stored the small ox-bow Long Lake is released when necessary through a series of channels into the headwaters of the Sunflower River.

The primary objective is to maintain a base flow in the Sunflower River of 122,000 cubic meters per day.

The well field

Location of production and observation wells.

The well field consists of:

• 11 high capacity production wells (13,000 m³/day).
• 7 observation wells

Topography

Surface topography.

Topography is typical of the flood plain of a large river; low relief, with the most obvious features being active and relict ox-bow lakes.

• Relief: between 47 and 52 m above mean sea level.

Land Use

Land use within the study area, 2003 National Agriculture Statistics Service, Cropland Datalayer.

Major crops:

• Cotton
• Soybeans

Hydrogeology

Based on drillers logs, the wells tap a shallow, confined aquifer of Quaternary age.

Geologic cross section from production well 02 to production well 10.

Geology:

Aquifer tests

Horizontal hydraulic conductivity determined from aquifer performance tests.

2 aquifer pump tests

• Test 1: Production Well 02
  • Time: 44 hours

• Test 2: Production Well 11
  • Time:42 hours

Solved for hydraulic conductivity and storativity using the Theis method.

• Average values for material properties were used in the model.

Transmissivity

Storativity

Numerical Model

Numerical model mesh.

Modeled using MODFLOW under GMS.

Spatial discretization:

• 6,494 active cells
• 152.4 m x 152.4 m cells
• used "convertible" cells to account for the possibility that the water table might drop beneath the top of the upper confining unit.

Temporal discretization:

• 61 stress periods (each represents 1 month)
• from April 1, 2001 to April 30, 2006
  • The YMD collected water-table levels in the study area semi-annually during this period.

Boundary Conditions

Boundary conditions, irrigation wells, observation wells, and well field production wells.

• Time variant specified head for Mississippi River boundary and SE boundary.
  • For the SE boundary we used data from semi-annual water-table survey.
• No flow at the NE and SW boundaries.
  • Model was oriented at an angle so that these boundaries would be approximately parallel to the regional groundwater flow.
• Recharge from the top boundary (5% of precipitation)
• No flow bottom boundary.

Top boundary

Monthly areal recharge during each stress period.

Recharge = 5% of monthly precipitation based on Krinitzsky and Wire (1964).

Mississippi River boundary

Groundwater level fluctuations in the alluvial aquifer in wells B012 (1.2 km from Mississippi River) and F005 (14.4 km from Mississippi River) compared to the elevation of the Mississippi River (Byrd, 2005).

Water levels in wells near the Mississippi River respond to the river's stage and fluctuate much more than wells further away.

Alluvial aquifer head versus Mississippi River elevation in well B012 show strong correlation.

Specified heads for each monthly stress period along the Mississippi River boundary.

We use a time-variant, specified head boundary.

• Monthly average stage interpolated from Mississippi River gauges and adjusted for the slope of the river.

Ox-bow lakes: General head cells

Elevation of general head cells representing the elevation of Moon Lake and Long Lake.

General head cells were used to represent the portion of the lakes in the model.

• Long Lake elevations linearly interpolated based on semi-annual measurements (April and October) by YMD.
• Moon Lake elevations extrapolated from daily measurement in 2006.

Irrigation wells

Agriculture irrigation wells and their irrigated area.

• 139 irrigation wells
• 6,400 m³/day cumulative discharge
  • 74 km² irrigated

• Model uses average values based on the YMD's annual agricultural use studies of values for cotton and soybeans (0.0026 m³/ m²/day

Initial conditions

Location of wells with driller’s logs or geophysical logs.

Initial water table elevations from the 2001 survey were interpolated onto the numerical mesh using ordinary kreiging in ArcGIS.

• Wells both in and outside of the model domain were used to get the best interpolation.
• Elevations of the top and bottom of the aquifer were interpolated in a similar method from the drillers' and geophysical logs.

Model Calibration

Observed versus simulated hydrographs

Simulated versus observed hydrographs for 3 wells closest to well field.

There is a good fit between simulated and observed water level data.

Aggregate observed versus simulated water levels

Simulated versus observed heads (correlation coefficient = 0.97).

Manual calibration for:

• hydraulic conductivity = 184 m/day
• specific storage coefficent = 2.2 x 10^-4 meter^-1
• specific yield = 0.32
• areal recharge = 5% of total monthly precipitation.

All hydraulic parameters were within reasonable range of previously reported Mississippi River Alluvial Aquifer properties.

RMS Error

Model error over time.

The change in the Root Mean Square error shows that the simulated heads converge to good agreement after the first year.

• After the first year, the average RMS was 0.37 m which is approximately 3.1% of the range of heads in the aquifer.
• Larger errors at the beginning of the simulation indicate that the model improved as it got away from the interpolated initial conditions. While Kreiging is as good a way as any to interpolate it does not account for the physics of the groundwater equations.

Residuals

Distribution of residuals.

The model produces residuals that are:

• normally distribution with a small positive bias, and,
• random

Randomness of residuals.

Sensitivity Analysis

Model error over time.

Specific model parameters were varied from 0.1 to 100 times their calibrated values to determine the sensitivity of the model to these inputs. Parameters considered:

• horizontal hydraulic conductivity
• areal recharge rate
• specific storage
• conductivity of general head cells beneath the lakes.

Modeling Results

Net fluxes

Modeled average annual net flow from April 1, 2002 to April 30, 2006.

Observations:

• Small net increase in storage over time period.
  • Only 0.01% increase per year (∆h = 0.0042 m/yr)
• Large net discharge to the Mississippi River
• Significant recharge from Long Lake

Monthly flow budget

Average monthly net flow from April 1, 2002 to April 30, 2006.

• Net recharge in winter and spring when there is high precipitation and high Mississippi River stage.
• Net discharge in summer and fall when precipitation is low and irrigation is highest.

Simulating well field operation

To determine the effects of the well field on the local and regional water table we performed a simulation with maximum expected pumping conditions.

• Pumping all wells at maximum for 6 months.
  • produces 143,000 m³/day
  • a more reasonable expectation would be for pumping through only 2 months
• Input conditions for October 1, 2002, through March 31, 2003.
  • average climatic conditions
  • low RMS error in model

Results: Drawdown

Simulated heads in the alluvial aquifer (meters msl) on March 31, 2003 after six months of continuous well field pumping compared to modeled heads without pumping.

Without pumping.

With 6 months of pumping.

Close view of well field.

Results: Recovery

Simulated drawdown in the production wells (meters msl) after six months of continuous pumping.

3 stages:

• days 0 to 60: Initial drawdown due to wells.
• days 60 to 190: Gradual increase in water levels.
• days 190 to 200: Rapid recovery after wells are turned off.

Induced recharge from the Mississippi River

Simulated flow from the Mississippi River to the alluvial aquifer induced by withdrawals from the well field.

Induced recharge from Long Lake

Simulated flow from the Long Lake to the alluvial aquifer induced by withdrawals from the well field.

Model parameters

Effect of proximity to the Mississippi River

Simulated heads in the alluvial aquifer (meters msl) that would result after six months of continuous pumping if the well field was located eight kilometers farther from the Mississippi River.

What is the influence of the Mississippi River on the well field?

After 6 months of pumping,

• 54,700 m³/day induced recharge from the Mississippi River.
• Flow in the Mississippi River is approximately 1.3 billion m³/day

Simulated drawdown in production well five at the well field’s current location compared to simulated drawdown in production well five if the well field was moved eight kilometers farther from the Mississippi River.

Note: The long recovery time that extends past the 365 day mark.

• This indicates that successive years of pumping will result in cumulative drawdown.

Effect of Long Lake

Simulated heads in the alluvial aquifer (meters msl) that would result after six months of continuous pumping if Long Lake was not located near the well field?

What effect does the water storage in Long Lake have on the impact of the well field?

After 6 months pumping:

• Induced recharge from Long Lake = 90,000 m³/day
• Estimated evaporation from Long Lake = 3,000 m³/day
• Leaving approximately 50,000 out of 143,000 m³/day for flow augmentation.
  • Note target flows in the Sunflower River are approx. 86,400 m³/day

Simulated drawdown in production well five with and without recharge from Long Lake.

Conclusions

Locations of proposed well fields in southern Mississippi.

1. The Sunflower River Flow Augmentation Project should have minimal effect on water table levels beyond 1 km of the pumping wells.
2. Proximity to the Mississippi River results in:
   • Induced recharge from the river.
   • Rapid recovery of the well field from pumping.
3. Long Lake severely restricts the expansion of the well field's cone of depression.
   • However, induced recharge from Long Lake reduces the amount of water available for flow augmentation (under simulated conditions) by over 60%.