WFS requests
From Intamap
Back to Proposed System Architecture
Contents |
[edit]
This is Dan's initial outline of what request the WFS might support
[edit]
Supported insert operations (i.e. what we can send to WFS):
- Insert observations (with data set ID / GML ID to make persistent) and also can include information on the sensor (how to pass to interpolation algorithm - MathML? Also should include sensor error characteristics?).
This is to a large extent now addressed by the intention to use SWE common and thus SensorML and Obs & Measurements - the question now is the role of GML?
[edit]
Supported interpolate requests:
- Specify algorithm (e.g. simple kriging / Automap / PPK / Spartan / Full Bayesian / ....) - this should be a string and a "Method not known" exception returned if the system does not implement the method. Probably there should be a default method if this is not specified?
- Specify optional parameters to the interpolation methods (with suitable defaults specified on the system if used in automatic mode) - i.e. first guesses, number of iterations, search radius ....?
- Specify which variables this is desired for? Multivariate?
- Return the hyper-parameters - for example the variogram / covariance parameters?
- Return meta-data / diagnostics about the algorithm employed in the automatic interpolation? Optional?
[edit]
Types of features for which we can request interpolated answers:
- GML rectified grid (a pretty much arbitrary grid format - results can be in a binary (i.e. compressed) format - specify the grid in request. Have a default.
- GML polygon (a closed polygon) - answer could be a grid within the polygon or an average over the polygon? - this should be specified with the request? If a grid is to be used, it must also be specified, as above.
- GML line (a polyline) - is this really needed? Possible use: find radiation levels along this route? Specify density of points along line?
- GML point. The only 'easy' one.
[edit]
What we can request interpolated answers to be (could be at a point, along a line, over an area, on a grid):
- mean - our best estimate
- mode - our most probable estimate (for a Gaussian these will be the same!)
- variance - marginal prediction variances at each location
- standard_deviation - square root of the above!
- covariance - joint covariance matrix for all prediction locations (note this could be huge!) - is this really useful - er can we compute it easily for all methods??
Are some of the above better simply phrased as requests for the moments of the distribution? This is more flexible, so the request could be getFeature [CenteredMoment(momentOrder)]?
- pdf - return the probability density function of the variables - how? Should this include the functional form of the pdf, e.g. in MathML?, mixture models? How?
- probability of exceedance - return for a given threshold the probability of exceeding the threshold - request passes threshold.
Again this is probably a specific case of the more general request to return the probability of a value lying in a defined range (including -infinity to +infinity) - thus the minimum and maximum values are specified.
- confidence interval - return a specified confidence interval for the (posterior) predictive distribution - request has the CI, e.g. [.05,.95].
- quantiles - this would be more general than confidence intervals since a range of quantiles of the distribution might be requested.
- histogram - return a histogram of the distribution - request specifies the number / location of bins. In grids is it a histogram for each point???
- samples - return samples from the joint interpolated (posterior) process - request passes the number of samples.
Please edit this page and add in any requests you think the web service should support, or comment on my foolish ideas!
Back to Proposed System Architecture
