Week 11: Spatial data uncertainty Flashcards
(23 cards)
All spatial data have some
Uncertainty
What determines if and what we can quantify uncertainty
Depending on how data was captured
Uncertainty has three components
- Error
- Randomness
- Vagueness
What is error in uncertainty
The known uncertainty due to systematic and human limitations
What is randomness in uncertainty
The component of uncertainty that despite minimised error cannot be modelled directly e.g minute variations in a surface
What is vagueness in uncertainty
This is uncertainty associated with a spatial or attribute concept e.g boundary or density
Two components error is split into
Accuracy and Precision
How is accuracy defined
How close the recorded data values are to true values or true description found in the real world
How is precision defined
Interpreted in terms of how exact data measurement and storage are
- The number of significant digits in a measurement
- Variability among repeated measurements
Precise data are not necessarily
Accurate and vice versa
Spatial data are affected by random and/or systematic errors that affect its
Accuracy - a combination of trueness and precision
Trueness characterises
- The closeness of the mean of the measurements to the actual (true) value, thus signaling systematic errors. 2. Typically expressed as arithmetic mean of residuals between measurements and reference values of better quality (e.g lidar referenced against surveyed control points)
Precision characterises
The closeness of agreement within individual results typically associated with random errors
Four sources of positional error
- Map projection, datum, and their parameters
- Primary measurement error - error determining location of features
- Instrument or human operator related - Secondary data acquisition errors
- Instrument (digitizer / scanner resolution)
- media (quality of source maps and photos
- Operator (registration / interpretation error) - Improper representation or conceptualisation of objects or phenomena
- Scale effects (0.5 mm resolution)
- Data model effects
Effect of scale: map positions usually accurate to about
1/2 mm at the scale of the map
Issue of uncertainty in analytical processes
Ecological fallacy
Ecological fallacy occurs when
It is inferred that the characteristics of an area can be applied to the individuals in that area
- e.g average income by meshblock (cannot assume that all individuals who reside in a specific meshblock have income levels equal to the meshblock average)
- e.g a town with high level of pollution and high level of cancer (cannot assume everyone in the town is likely to get cancer
The modifiable areal unit problem occurs when
Arbitrary boundaries are used to represent phenomena that vary continuously over space
Two effects of the modifiable areal unit problem
- Scale effect - variation in results when the same data is grouped at different levels of spatial resolution
- Zoning effect - variation in results by changing the boundaries of zones at a given scale
Error propogation
- Errors in one data set are carried through to other data sets created from it
- Widely ignored in GIS
Cascading error
Errors in one or more dat sets are propagated repeatedly into new layers through multipe data processing operations (overlay, feature selection, etc.)
Effect of the combination of propagating and cascading errors can be
Additive or multiplicative and are difficult to predict in advance
Six ways of managing error and uncertainty
- Acknowledge inevitable uncertainty in data
- Do not assume that data represents the truth - Undertake error assessment if possible
- Ensure all datasets are documented (metadata)
- Inform consumers of analysis outputs of degree of uncertainty
- Establish data quality and accuracy standards (e.g LINZ lidar standard)
- Sensitivity analysis to ascertain nature of uncertainty in analysis
