Capstone: climate adaptation and the Kestrel Creek master plan

The master plan comes together in this lesson. Its works will last 50 to 100 years: a culvert designed now will see the climate of the 2080s, and a dam spillway longer still. Anika's final task for the team is to test every recommendation against a changing climate, rank the projects for the city's capital program, and write the plan so that the next engineer can see what was assumed and why. The lesson treats climate adaptation as a design discipline rather than a forecast: the climate cannot be predicted precisely, but the designs can be made robust to a plausible range.

Nonstationarity

Classical hydrology assumes stationarity: that the statistics of rainfall and floods do not change over time, so that the past record describes the future. A warming climate breaks that assumption. Warmer air holds more water vapor, about 7 % more per degree Celsius by the Clausius-Clapeyron relation, and observations and models show the most intense short rainfall bursts increasing at roughly that rate in many regions, sometimes faster. Snow-dominated regions see earlier melt, more winter rain and changed spring floods. Low flows may fall where summers lengthen and evapotranspiration rises. Lake Superior's levels respond to the whole basin's balance and have ranged widely within the past few decades.

Design responses fall into three groups. Climate-informed design adjusts the design rainfall or flow with change factors from climate projections. Robust design chooses works that perform acceptably across a range of futures, even if not optimally in any one. Adaptive design builds now for a moderate future but leaves room to upgrade, for example by reserving land, oversizing foundations or designing a berm that can be raised, with triggers that say when to act.

US guidance includes NOAA Atlas 14, whose Volume 8 covered Minnesota and the Midwest in 2013, and its successor NOAA Atlas 15, designed to include climate trends and projected change, with a pilot in 2024 and preliminary nationwide estimates scheduled for review in 2026. FHWA's Highways in the River Environment (HEC-17), in its second edition, sets out risk-based and climate-informed approaches for highway hydraulics, and its vulnerability assessment framework guides transportation agencies. Minnesota's state agencies publish climate projections and resilience guidance for local governments.

Risk over the design life

A return period describes a probability per year, not a schedule. The chance that the \(T\)-year event is equalled or exceeded at least once in an \(n\)-year design life is

A culvert designed for the 100-year flood with a 75-year life has better than even odds of seeing its design flood. That is not a design failure, since the design flood is supposed to be survivable with freeboard, but it is a reason to check performance above the design flood and to make that performance graceful: overtopping of the road rather than washout of the embankment, controlled flooding of a yard rather than of buildings.

# Climate adaptation for the Kestrel Creek master plan: design-life risk, rainfall change and priorities
UNITS = "US"
SI = UNITS == "SI"
def risk(T, n): return 1 - (1 - 1 / T) ** n
print("Chance that the design event is equalled or exceeded at least once during the design life")
for T in (25, 100, 500):
    print(f"  T = {T:3d} years: 25 years {risk(T, 25):5.1%}, 50 years {risk(T, 50):5.1%}, 75 years {risk(T, 75):5.1%}")
# rainfall scaling: about 7 % more short-duration intensity per degree C of warming (an assumption to test)
for dT in (1.5, 2.5, 3.5):
    print(f"Warming {dT} C: rainfall factor {1.07 ** dT:.2f}")
# runoff amplification with the curve number method (CN 75, 100-year 24-hour depth from CIV-370)
P = 95.3 if SI else 3.75
CN = 75
S = (25400 / CN - 254) if SI else (1000 / CN - 10)
def runoff(p): return (p - 0.2 * S) ** 2 / (p + 0.8 * S) if p > 0.2 * S else 0.0
q0 = runoff(P)
for fac in (1.10, 1.15, 1.27):
    print(f"  rainfall x {fac:.2f}: runoff {runoff(P*fac):.1f} {'mm' if SI else 'in.'} against {q0:.1f} -> x {runoff(P*fac)/q0:.2f}")
# master plan priorities: risk reduced per dollar, with a minimum score for the environment
projects = [("culvert 4.5 x 2.4 m with fish passage", 1650, 8.5, 9),
            ("Kestrel Lake dam spillway widening", 820, 7.0, 4),
            ("Hillside Road bridge pier scour protection", 210, 5.5, 5),
            ("operations site LID retrofit", 480, 3.5, 8),
            ("floodplain buyout of two houses", 900, 4.0, 7),
            ("two supply wells for drought backup", 560, 6.0, 5)]
print("\nproject                                     cost ($k)  risk score  env score  score per $M")
for name, cost, rk, env in sorted(projects, key=lambda p: -(p[2] + 0.5 * p[3]) / p[1]):
    print(f"  {name:42s} {cost:6,d}     {rk:4.1f}       {env:3d}       {(rk + 0.5 * env) / cost * 1000:5.1f}")

Chance that the design event is equalled or exceeded at least once during the design life
  T =  25 years: 25 years 64.0%, 50 years 87.0%, 75 years 95.3%
  T = 100 years: 25 years 22.2%, 50 years 39.5%, 75 years 52.9%
  T = 500 years: 25 years  4.9%, 50 years  9.5%, 75 years 13.9%
Warming 1.5 C: rainfall factor 1.11
Warming 2.5 C: rainfall factor 1.18
Warming 3.5 C: rainfall factor 1.27
  rainfall x 1.10: runoff 1.8 in. against 1.5 -> x 1.19
  rainfall x 1.15: runoff 1.9 in. against 1.5 -> x 1.29
  rainfall x 1.27: runoff 2.3 in. against 1.5 -> x 1.52

project                                     cost ($k)  risk score  env score  score per $M
  Hillside Road bridge pier scour protection    210      5.5         5        38.1
  operations site LID retrofit                  480      3.5         8        15.6
  two supply wells for drought backup           560      6.0         5        15.2
  Kestrel Lake dam spillway widening            820      7.0         4        11.0
  floodplain buyout of two houses               900      4.0         7         8.3
  culvert 4.5 x 2.4 m with fish passage       1,650      8.5         9         7.9

Reading the results

The design-life table shows a 53 % chance that the 100-year flood will be exceeded at least once in 75 years, and a 14 % chance even for the 500-year flood. Applying 7 % per degree Celsius to warming of 1.5 to 3.5 °C (2.7 to 6.3 °F) gives rainfall factors of 1.11 to 1.27. Runoff amplifies the change: with the curve number method on the 100-year 24-hour rainfall of 3.75 in., a 15 % increase in rainfall raises runoff by 29 %, and a 27 % increase raises it by 52 %. That amplification is why CIV-370's climate-check flow of 850 cfs, a third higher than the 640 cfs 100-year flow, is a sensible test for the creek crossings.

Testing each recommendation

The table shows the plan's character. Most works are sized for today's standards but checked against a future flow, and almost all keep an option open: space for a fuse plug at the dam, land beside the pond, a third well site, a berm whose crest can be raised. Those options cost little now and make the plan adaptive. The one exception is the culvert, which is hard to enlarge later, so its size is set by the climate check, not by the present 100-year flow alone.

Ranking the projects

The city cannot fund everything at once. The team scores each project on the risk it reduces and on its environmental benefit, and divides by cost to rank them. The code's ranking puts the Hillside Road scour protection first, a small, cheap project that removes a serious risk to the only access road, followed by the stormwater retrofit and the wells. The culvert comes last per dollar, because it is expensive.

A pure ranking would be misleading. The culvert must be replaced anyway because of its condition and its failure as a fish barrier, and lesson 10 showed that the extra cost of the larger size pays for itself in avoided flood damage. The plan therefore separates mandatory projects, those driven by regulation, condition or safety, from discretionary ones, and ranks only the latter by value for money. Engineers should present scoring as an aid to judgment, show the scores and weights openly, and test whether the order changes when the weights change.

The master plan report

The plan's report records the problem, the data, the assumptions, the alternatives considered and the reasons for the recommendations, so that the city, the agencies and future engineers can follow them. Its structure for Kestrel Creek is: objectives and engagement; existing conditions, with the water balance, the flow regime, the crossings, the dam, the floodplain and the stream's condition; future conditions, including growth and climate; the alternatives for each element and their evaluation; the recommended plan with costs and priorities; the approvals each project needs; monitoring, including streamflow, piezometers, the dam's instruments, the bridge's bed and the creek's fish; and the triggers that would prompt the next review.

The approvals list includes an amended water appropriation permit for the co-op and the wells, DNR dam safety and public waters work permits for the spillway and the creek works, a Section 404 permit from the Corps for fill in the creek, the MPCA construction stormwater permit, a conditional letter of map revision from FEMA for the culvert's effect on flood levels, and the city's floodplain ordinance for the berm. The plan's environmental documentation supports any state environmental review that the larger projects trigger.

Monitoring and review

An adaptive plan is only adaptive if someone watches. The team recommends a stream gauge at Ridgeway Road, which also improves future flood frequency estimates; a rain gauge at the operations center; piezometers between the wells and the lift station with trigger levels; the dam's piezometers and seepage weir; a bed survey at the Hillside Road bridge each spring; and fish surveys above and below the new crossing to confirm that brook trout are moving through it. The plan is reviewed every ten years, or sooner if a trigger is reached: a flood exceeding the climate-check flow, a drought deeper than the design drought, a change in standards, or new climate projections that move the design values materially.

Practical points

• Test every design against a plausible future, not only the past record.
• Express risk over the design life; a 100-year design is likely to be exceeded in a long life.
• Keep options open at low cost: land, foundations, crest heights that can be raised.
• Separate mandatory from discretionary projects before ranking by value for money.
• Write down assumptions, monitoring and triggers so that the plan can adapt.