FAQ

1.1 What is INTAMAP?

INTAMAP stands for Interoperability and Automated Mapping. The main objective of the project (2006-2009) was to develop web services for the real-time mapping of critical environmental variables using advanced geostatistics. INTAMAP is based on the latest developments made in interoperable web based services. More info can be found at www.intamap.org and in the wiki http://wiki.intamap.org

INTAMAP is a project funded by the European Commission (DG INFSO, FP6)

1.2 What is WPS (Web Processing Service)

The Open Geospatial Consortium's (OGC) Web Processing Service (WPS) is a Web Service that takes a defined set of inputs, applies a calculation, and produces a defined set of outputs. WPSs are designed to work with spatially referenced data but can be used with other kinds of data. A WPS is a generic Web Service in that the specific processes provided by a WPS implementation are defined by the owner of that implementation.

More on the OGC WPS documentation page

1.3 Ordinary Kriging, psgp or Copula Kriging: what is the best for me ?

In a nutshell, the following decision tree can be used:

If the dataset has no major outliers, if its size it not too large (< 1000 points) and the grid is not exceeding a resolution of 100 x 100 (= 10 000 points to be estimated) then Ordinary Kriging (Automap) should be fine. Larger datasets can be quickly processed using psgp although this will have some cost in terms of average errors. In the presence of outliers, copula kriging would be recommended. No solution has been found so far for very fast geostatistical interpolation in the presence of hot spots.

1.4 Can I have my own interpolation/cross-validation server?

INTAMAP is 100% Open Source and everything is located in the projects SVN https://intamap.svn.sourceforge.net/svnroot/intamap

1.5 Where can I find servers with the intamap interpolation service installed

Project partners have installed and maintain the service on the following servers:
http://intamap.uni-muenster.de:8080/intamap/WebProcessingService
http://intamap.geo.uu.nl:8180/intamap/WebProcessingService
http://intamap.aston.uc.uk:8080/intamap/WebProcessingService
You can test the service by sending xml-requests through one of the test-pages, e.g.:
http://intamap.geo.uu.nl:8180/intamap/test.html

1.6 What are EPSG-codes

EPSG-codes were defined by the European Petroleum Survey Group as an official list of codes for different map projections and datums. An overview can be found on Spatial Reference. Some commonly used epsg codes are:

• EPSG: 3035 Single CRS for all Europe. Used for statistical mapping at all scales and other purposes where true area representation is required.
• EPSG: 4326 Lattitude/Longitude coordinates - WGS 84. Used by the GPS satellite navigation system and for NATO military geodetic surveying.
• EPSG: 28992 Dutch projection, used for the Meuse data set, often used for examples

2.1 General

2.2.1 How do I install the WPS

Installation instructions are given here.

2.3.1 How do I use the interpolation service ?!

Web Services work with XML. Hence, your data sets (typically ID, X-Coord, Y-Coord, Z value) need to be converted in an XML scheme. The services provided by INTAMAP use Observations and Measurements schema (Obs & Meas info) for data description and GML (info) for domain description.

It is also possible to use Sensor Observation Service as a data source; in this case the data input is just an URL

2.3.2 I need more technical information on what I can submit for interpolation ?!

Please see INTAMAP's wiki and "What can you send to the WPS"

2.3.3 How do I define my interpolation domain?

The domain of interpolation can be set through the WPS request. The domain can be defined using the following GML geometries: Point, Polygon, MultiPoint, MultiPolygon and RectifiedGrid. Please see INTAMAP's wiki and "What can you send to the WPS"

If you do not define a region of interest where interpolated values are desired, a 10 000 point grid (100 x 100) will be set by default using the external points as boundaries.

2.3.4 Where can I find an example on how to use the interpolation service?

The page send an XML request gives a very clear example.

If you prefer you can build your own request: An example of a request made to the interpolation service can be downloaded from the project's SVN. The Meuse data set (link) is used in the example.
The downloaded file contains a WPS request that contains the Meuse data and a basic interpolation grid. The WPS request sets Automap as the default interpolator. It is possible to send the request using the wget command (windows users see here) To send a request:

#>wget -nv -q -O - --postfile=meuse-automap-grid.xml http://remwps2.jrc.ec.europa.eu/intamap/WebProcessingService

2.3.5 How do I define parameters for kriging interpolation?

The major objective of INTAMAP is to fully automate the mapping process. Hence, all parameters are either preset or their choice has been optimized. If you want more control then you need to use the interactive versions of the intamap packages - see Download / Installation section.

3.1.1 What is Ordinary Kriging?

Ordinary Kriging is the core interpolation function in geostatistics and is designed to estimate values of a continuous variable (e.g. elevation, temperature, rainfall, environmental radioactivity, ...) at an unobserved location using observations taken from nearby locations. The automation of the use Ordinary Kriging has been implemented in Automap, a r-cran package that contains several functions that will automatically fit a model of the spatial correlation (the semivariogram) in a spatial dataset. Automap works on top of gstat and sp

Detail information can be found on the following publications/links:

Hiemstra, P.H., Pebesma, E.J., Twenhöfel, C.J.W and G.B.M. Heuvelink (2009). Real-time automatic interpolation of ambient gamma dose rates from the Dutch Radioactivity Monitoring Network. Accepted for publication at Computers & Geosciences (Elsevier).

Hiemstra, P.H., Pebesma, E.J., Twenhöfel, C.J.W and G.B.M. Heuvelink (2008). Automatic real-time interpolation of radiation hazards: a prototype and system architecture considerations. International Journal of Spatial Data Infrastructures Research, vol 3, p 58-72 (http://ijsdir.jrc.it/)

3.1.2 What is psgp?

PSPG stands for Projected Sequential Gaussian Processes. The PSGP is designed for dealing quickly with very large datasets. The method discards data by sub-sampling the original observations in a way that the sub-sample will contain sufficient information for a "correct" interpolation. The PSGP method also uses BLAS and ATLAS libraries to optimize matrix inversions and calculations. The PSGP method also has the capability of using uncertainly associated with measurements. More details can be found here.

3.1.3 What is Copula Kriging ?

In the common Ordinary Kriging approach, the analysed variable is considered to have a normal distribution. Environmental data often contain a few extreme values called "hot spots" that are affecting the Gaussian distribution of the data. The copula method is a multivariable way of analysing data containing such "hot spots". The method is very recent and more details can be found only in a few papers for the moment:

Kazianka, H and J. Pilz . geoENV VII - Geostatistics for Environmental Applications Series: Quantitative Geology and Geostatistics, Vol. 16 Atkinson, P.M.; Lloyd, C.D. (Eds.). Springer New York

Kazianka, H and J. Pilz (2009) Bayesian spatial modeling and interpolation using copulas. In: "Proceedings of the StatGIS09: GeoInformatics for Environmental Surveillance", June 17-19, 2009 Milos, Greece

3.1.4 What is IDW (Inverse Distance Weighted) Interpolation ?

The Inverse Distance Weighted interpolator is a standard deterministic interpolator that is easy and fast. IDW can be used to compare the benefits of using more advanced functions like those proposed in INTAMAP. The exponent value of IDW, traditionally set to 2 (Inverse Squared Distances) is here selected automatically on the basis of a minimum Root Mean Square Error found from cross-validation (CV). This CV is by default 5-fold CV, i.e. that the data set is split in 5 folds, and the process at the observation locations of each fold are interpolated from the remaining 4 folds. This random splitting in data sets can cause that the best fitted value for the exponent can vary for the same data set.

3.2.1 I have a problem with rJava when installing intamapInteractive

This is usually caused when R cannot find your Java installation. A possible solution is usually to run: R CMD javareconf It might be necessary to run this command as administrator of your computer.

3.3.1 Why/when do I have to regenerate the time models

The time models supplied in the data set are valid for a certain installation of R and packages on a particular computer. Computation time will differ on different computers, depending e.g. on the CPU, different libraries installed (e.g. BLAS/ATLAS), and on the version of different R-packages. Although most updates of R-packages will not lead to performance changes of the interpolation procedures, it is still recommended to regenerate the time models on a regular basis (at least every 6 months).

3.3.2 Is the estimate of the prediction time exact?

No, the estimate is an estimate. It can be relatively accurate for some methods (IDW/automap), whereas the uncertainty is larger for psgp and copula. This is for three reasons. First of all are the time models generated from a simulated data set, different observations and configurations can influence the time. Most important is that psgp and copula methods estimate the parameters through iterative search procedures. Different paths can be used for finding the parameters, and different data sets can influence the convergence of the optimization routines. The last cause of differences is that most methods can fit a range of models to the data sets, and the choice of model will also influence the prediction time.

3.4.1 Can I start Rserve from every directory?

No, it is important that you start Rserve in the same directory where you started R when you installed Rserve. This assures that you also load the correct .First-function loading the psgp and/or intamap packages.

3.4.2 I get an error message about missing coordinates

If your error message looks something like: Error in coordinates(datapoints) = ~x + y :
could not find function "coordinates<-" it is most likely that the intamap package has not been loaded. There are two common causes for this:

• The psgp/intamap package is not installed in the directory referred to as Rlibs in the Rserve installation guide.
• Rserve was not started in the same directory where R was started to generate the .First-function. See the Rserve installation guide and the FAQ-question above this.

4.1 General

4.1.1 How do I check the quality of my interpolation results?

"Quality control" is done by running a crossvalidation service on the same dataset and parameters used for interpolation. See cross-validation

4.1.2 Where can I find a running service ?

In the following URL: http://remwps2.jrc.ec.europa.eu/cgi-bin/wps.py

4.1.3 Where can I find an example on how to use the cross-validation service?

Try the project's SNV and the cross-validation/request folder: link

4.1.4 How do I send a request?

One possibiliy is using wget, see question 2.3.4

4.1.4 What outputs do I get back from the cross-validation service?

You will receive statistical information (Root Mean Square Errors = RMSE, Mean Errors (ME), Mean Absolute Errors (MAE), Pearson's correlation coefficient computed on the estimated values against the true values) as well as graphical information, typically a map showing a proportional symbol plots of the residuals, a histogram of the residuals, a variogram of the residuals (ideally, residuals should not be correlated). An example of output can be found in the wikipedia (link)

5.1.1 What clients can use the INTAMAP system?

There are several clients that can use the INTAMAP system - these are described in the Using INTAMAP section.

6.1 Who shall I contact in case of problems???

The project itself is finished, but many of the partners will continue working on related issues. There is a mailing list for questions and further developments:
intamap-info at lists.sourceforge.net
If you want to join this mailing list, please go to
https://lists.sourceforge.net/lists/listinfo/intamap-info
For contact information for specific partners, see the contact page: http://intamap.org/contact.php

7. Bugs and known problems