Hello, All I have posted this message to the gstat-info mailing list - apologize for the cross-posting.
I am processing oceanographic datasets from NWW3 (NOAA) model in order to obtain subsets of data refined both in time and in space. The main goal is to interpolate the original spatio-temporal grid to smaller space and time lags without having to accopolate another wave-propagation model. I have considered using traditional interpolation methods (IDW,...) but I decided to use simple kriging, which enables me to incorporate space-time dependencies in the interpolation process. Nevertheless, I have bumped into two awkward problems. The first one may be quite trivial but I haven't figured it out so far. Some of the data on the NWW3 model is related with wavesets direction. This data is presented in polar coordinates ranging from 0-360 degrees, which poses a serious problem on mean value calculation - a fundamental process on variogram estimation. For example, if one accounts for two wavesets directions in two different points, one with direction 10º and the other with direction 30º, the mean direction is 20º - so far, so good... But, when considering two wavesets with directions 5º and 355º, the naive mean will be 180º which is extremely different from the desired 0º value. A solution for this problem is the separate averaging of sine and cosine values and reverting the process using the arctg function. This is the basic maths approach but how do I transfer it to variogram estimation? Should I decompose my directions field in two other, sine and consine, evaluate the anisotropy of each one of these and hope to have the same ellipse defined from two dijunct variables? Well, I am just starting in geo-statistics so... much help is needed to solve this problem... The other problem is an operational one. I am using gstat automatic modelling capabilities to define my ellipsoid axis, calculating variograms of four different horizontal directions, which works OK most of the time. However, sometimes it doesn't and I end up without an orthogonal pair of axis and hours of useless work. The question is: how can I speed up the process of discovering the main/minor directions of my ellipses without the exhaustive calculation of (too) many semivariograms - i.e., many directions? I am feeding gstat with some hundreds of thousands of lines of data each time I ask it to calculate a variogram - which is a slow process even when using a dual pentium, 2gb ram, 500gb disk box... I would deeply appreciate your help on these issues. Best regards, Nelson Silvestre MSc GIS student Instituto Superior Técnico - UTL, Technical University Lisbon + + To post a message to the list, send it to [email protected] + To unsubscribe, send email to majordomo@ jrc.it with no subject and "unsubscribe ai-geostats" in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list + As a general service to list users, please remember to post a summary of any useful responses to your questions. + Support to the forum can be found at http://www.ai-geostats.org/
