Supporting Information
Environmental Science & Technology
Reassessing Hypoxia Forecasts for the Gulf of Mexico
Donald Scavia* and Kristina A. Donnelly
*Corresponding Author
Contact information for Corresponding Author:
Phone: 734-615-4860
Fax: 734-763-8965
E-mail: [email protected]
7 Total Pages
Figure S1 – Nutrient Load Regressions
Figure S2 – Hypoxia Area-Length Regression
Figure S3 – Model Calibration Results
Figure S4 – Nutrient Load – Hypoxia response curves
Model description - The model is a variant on the Streeter-Phelps dissolved oxygen
model, commonly applied to river and estuarine systems. It simulates the concentration of
decomposing organic matter (B) in oxygen equivalents (i.e., BOD, mg L-1) and the
dissolved oxygen deficit from saturation (D, mg L-1) at increasing distances downstream
from two point sources of loads; namely, the Mississippi and Atchafalaya Rivers. The
mass balance equations are as follows:
dB/dt = -vdB/dx – aB
dD/dt = -vdD/dx + aB – bD
where t is time (d), x (km) is the distance from the point source, a is a first-order rate
constant for organic matter decomposition (d-1), and b is a first-order rate constant for
vertical oxygen flux (d-1). The calibration term, v (km d-1), takes into account both the net
westward advection (as in the original Streeter-Phelps model) and all other uncertainties
in model structure, parameters, and drivers.
Assuming no upstream oxygen deficit, and ignoring longitudinal dispersion, the model’s
stead state solution is:
B = B0e-ax/v
D = [a / (b–a) B0(e-ax/v – e-bx/v)]
The model assumes that subpycnocline oxygen consumption can be approximated as
first-order decay of organic matter, and that oxygen flux across the pycnocline is firstorder and proportional to the oxygen deficit (here defined as the difference between
subpycnocline concentrations and 7 mgL-1, the saturation value at typical Gulf surface
water temperatures and salinity).
We also assumed that there is a relationship between the observed mid-summer hypoxia
and that predicted at steady state, and that movement of sub-pycnocline water west of the
Mississippi and Atchafalaya rivers could be modeled as a one-dimensional flow. In
reality, the spring and early summer production of organic matter, its subsequent rapid
westward movement along the surface, and its settling into the slower-moving bottom
waters is a dynamic transition into the summer and early fall development of hypoxia.
However, our ability to accurately simulate the year-to-year variability helps validate the
use of point-source approximations, our parameterization, and steady state assumption.
This assumption is also supported by the fact that the Gulf is highly stratified in the
summer, which causes the subpycnocline water to be isolated from the unpredictable
dynamics of the surface water flow, and the fact that most of the interannual variability in
hypoxic area can be explained by variation in its length (Figure S2). Expansion beyond
these simplifications with, for example, multi-layered, time-dependent detailed
ecosystem models may add further insight into the transition periods; however, they also
require significantly more parameter estimation, calibration, and field data.
Hypoxia characteristics and data -- Relatively slow, coastally-constrained flow,
coupled with other biogeochemical and physical processes, produces the characteristic
shape of the hypoxic zone. These hypoxic areas typically extend westward from the
mouths of the Mississippi and Atchafalaya rivers as much as 500-600 km, past the Texas
border. In contrast to their length, they are typically only 30-60 km wide and less than 10
meters thick. Hypoxic zone lengths and areas were derived from contours of bottomwater oxygen concentration less than 2 mg L-1 generated from field data using the
Kriging interpolation method of Surfer® 7. Distances along the length of the hypoxic
area(s) were measured using ArcView® software on imported Surfer®-generated
contours. The characteristic shape of the hypoxic area, supporting the strong dependence
on long-shore transport, leads to a significant simple linear regression of area with length
(slope = 38.9; r2 = 0.82, Figure S2). This strong correlation and the relatively slow, lowfrequency currents observed in the bottom waters justify our assumption of a onedimensional analysis.
TKN (MT/Month)
a.
y = 11.971x 0.7148
R2 = 0.745
1,000,000
100,000
10,000
1,000
1,000
10,000
100,000
1,000,000
b.
MELV TN (MT/Month)
NO3 (MT/Month)
y = 0.8039x 0.9258
R2 = 0.9307
100,000
10,000
1,000
10,000
100,000
1,000,000
STFR TN (MT/Month)
y = 0.2167x 1.0912
R2 = 0.8171
c.
TP (MT/Month)
100,000
10,000
1,000
100
1,000
10,000
100,000
d.
MELV TP (MT/Month)
Discharge (m 3/s)
y = 0.5239x0.9678
R2 = 0.7582
100,000
10,000
1,000
100
1,000
10,000
STFR TP (MT/Month)
Figure S1. Regressions used to reconstruct historical TN and TP loads.
100,000
Hypoxic Area (km2)
25,000
y = 38.835x
20,000
2
R = 0.8178
15,000
10,000
5,000
0
0
200
400
Hypoxic Length (km)
600
Simulated Hypoxia (km 2)
Figure S2. Relationship between hypoxic region length and area.
30000
25000
20000
15000
10000
5000
0
0
5000
10000
15000
20000
25000
30000
Oberved Hypoxia (km2)
Figure S3. Calibration results for TN-driven (squares) and TP-driven (diamonds) models.
Solid line represents perfect agreement
Hypoxic Area (km 2)
28000
24000
20000
16000
12000
8000
4000
0
0
2000
4000
6000
8000
Total N Load (MT N/Day)
10000
2
Hypoxic Area (km )
28000
24000
20000
16000
12000
8000
4000
0
0
200
400
600
Total P Load (MT P/Day)
800
Figure S4. Response of Gulf hypoxia to reductions in TN (a, top) and TP (b, bottom)
loads. Model output is represented as mean ± standard deviations of 1000 Monte Carlo
simulations. Horizontal black line represents action plan goal for hypoxia. Vertical lines
represent 1980-1996 mean loads (dashed; far right), and mean ± one standard deviation
loads required to achieve the goal.