Observation uncertainty
19 important questions on Observation uncertainty
Give the water balance for a headwater catchment and a polder error?
Polder: seepage, artificial drainage as part of outgoing Q, surface water supply as incoming. See slide!
How does Bouaziz try to estimate IGF?
- Budyko analysis flipped
- Hydrological model
- Actual ET estimates
Explain the Budyko curve and what it is used for?
Etact = ETpot, so catchment is quite wet. Line has slope of 1
ETact/P = 1 : all P evaporates
Wetter catchments left, arid climates right.
Real catchments are somewhere along the curve.
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
What are important radar measurement errors (used for estimating P) and how can they generally be corrected?
- Overshoosting: measuring above clouds
- Attenuation: storm close to radar blocking signal beyond
- Ground clutter: mountain/buildings close influencing the signal
Corrections:
- Raingauges
- Statistical filters (useful for ground clutter) and physical corrections
Kinda skipped min 26 ish (rainfall slide)
Slide 17: How do rainfall errors propagate in models through various measurement techniques (scatter plots)? See min 27-32 ish
x-axis: gauge Twente/unadjusted radar/etc.
y-axis: local rain gauge=base run
How do rainfall errors influence the water balance (histograms)?
Microwave linked data: overestimates in several months
What are the three main error types and how do they influence the model outcomes?
- Biases (unadjusted radar): amplified when propagated through the hydrological system
- Timing errors (gauge outside catchment): attenuated when propagated through the hydro system (thus errors in individual hours in P, but in Q it is less of a problem)
- Seasonally varying errors (microwave link data): affect dynamics of simulated catchment water balance
How is Q often measured?
How do you obtain a rating curve?
- Measure Q several times, together with h
- To get Q: measure v, multiply with A
- Measure v with propellor or ADCP or other
- Interpolate/extrapolate (h,Q)-plot between observations
How does the ADCP work?
- Send an accoustic signal in the water and solid particles reflect the accoustic signal. Hereby derived v from doppler shift (reflection flowing particles)
- Sometimes fixed (horizontal) ADCP instead of rating curve (tides, backwater)
- Accuracy: often assumed to be <5%, but in practice worse
- Difficulties: flow through floodplains (floods), edges missing, compass calibration, moving bed
How does rating curve uncertainty propagate, based on the 3 plots? (slide 29)
Black pointing lines are measurements (?).
Red: rating curve
Min 59
Bottom left:
Q99: discahrge that is exceeded 99% of the time (low Q)
Q/Qactual discharge
Q0.01: Q exceeded only 0.01% of time, thus very high Q
Relative error becomes smaller to the more common Q values
Bottom right:
Missed it, min 61
Section: Examining Q time series: error or artefact?
Peculiarities in Q time series:
- Peculiarities: what to do with it?
- What caused it? > Crime Scene Investigation
Error or artefact (case 2): Imagine you have weir and Q is measured upstream a little. Weir is used as a flow measurement structure.
Two TS: green dashed is reference, blue: Qdatasets.
Blue line is higher overall then the green ref line
What could have happened here?
- Wrong catchment area used for conversion m3/s --> mm3/h
- Fix it: Water balance analysis
Error or artefact (case 5): Imagine you have weir and Q is measured upstream a little. Weir is used as a flow measurement structure.
Two TS: green dashed is reference, blue: Qdatasets.
Blue perfect cover, but starts overestimating a large rainfall event
What could have happened here?
- Weir was moved upa nd down, but weir height not registered
- Wrong Q-h relation used --> reconstruct or remove
Error or artefact (case 6): Imagine you have weir and Q is measured upstream a little. Weir is used as a flow measurement structure.
Two TS: green dashed is reference, blue: Qdatasets.
Blue very erratic but accurate signal
What could have happened here?
- Target level from water authoraties on which water is managed. If settings are too precise, water levels change all the time. Weir automatically adjusted after small water lvl variation
- Real data, but moving average useful for modelling purposes
Error or artefact (case 7): Imagine you have weir and Q is measured upstream a little. Weir is used as a flow measurement structure.
Two TS: green dashed is reference, blue: Qdatasets.
Blue overshoots one specific, max peak
What could have happened here?
- Weir blocked (by eg debris due to intense P) or submerged (crest of downwater is higher than weir and submerges it)
- Wrong Q-h relation used > reconstruct or remove
Error or artefact (case 8): Imagine you have weir and Q is measured upstream a little. Weir is used as a flow measurement structure.
Two TS: green dashed is reference, blue: Qdatasets.
Blue is same TS but shifted a little to the left (earlier in time).
What could have happened here?
- Timezone/clock settings are wrong, eg winter/summertime (this happens often!!)
- Shift entire series (crosscorrelation function may help)
What are 4 take home messages from Claudia on uncertainty of observations and modelling?
- Always set up the water balance first!!
- Critically evaluate observations before putting them in a model
- Look for causes of errors
- Think (or test) how errors propagate
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding
![Trustpilot Logo](/images/design_2020/lp/clp-trustpilot-en.png)