Excellent lecture! I have used Kriging for more than a decade but did not have the insights that you presented in your 36 min lecture. Also, the excel spreadsheet very useful in getting an intuitive understanding of the Kriging method. Much appreciated!
The right amount of math to make it applicable, but not so much that it remove the focus on what is happening. Thanks a lot for putting this quality content up on youtube!
Thank you Mathias for the feedback. I love math, but I love accessible education even more! I'm always trying to balance! I'm glad to hear that the content is useful!
Great lecture!!!. (correct me if I am wrong) the main assumption here is to have a signal noise representation of the signal f=m+e, then for predicting a new signal f*=m*+e*, kriging assumes that e*=sum_i a_i e_i = sum_i a_i (f_i-m_i) which gives the equation f*=sum_i a_i f_i + m*- sum_i a_im_i. Then weights a_i are founded by minimizing the variance with constraint sum_i a_i=1. (1) All this start by modeling e*=sum_i a_i e_i right?, 2) If we assume that f's are realization of Gaussian process, then f* can be estimated using equation 2.19 or 2.25 from this book www.gaussianprocess.org/gpml/chapters/RW.pdf i.e., the approach you are using wil be equivalent (equal) than the showed in that book, right?). Thank for sharing this video
Yes it is awesome! I watched all your lecture series on Data Analytics and Geostatistics from 1a Data Analytics Reboot: Statistics Concepts to the last. As a graduate student in environmental science with biological sciences as undergrad who has limited statistics and mathematical background, your videos and teaching materials are very helpful as they are easier to understand than papers/books which are also difficult to obtain. Thank you very much for your generosity in sharing your expertise to wannabe geostatisticians: your lecture materials (PDF format), videos, program codes in R and Python. I'm experienced in R but not in Python, but your Python workflows motivated me and now I can program in Python as well!
@@renatojrfolledo5728, thank you for sharing this. Now I'm totally stoked. This is why I do it! We are all one scientific community and more data analytics will improve science everywhere! Thank you, Michael
Thanks a lot for the great lecture. Learning from this lecture that Kriging accounts for distance (i.e., in the lecture, increase/decrease weights if a sampling point is closer/further away from the unknown location) when assigning weights, would you still recommend to perform data declustering and calculate the weights for data? I just wonder if account for data closeness twice would be redundant? Thank you very much in advance!
That is a great idea, Nazmul. I mentioned universal kriging in class for completeness to expand on the idea that common kriging variants are driven by various stationarity assumptions. I'll put together some content on this. Aside, given my geoscience-oriented engineering background I generally prefer mapped trend models. Thank you, Michael
Thank you for the great lecture i have a question about the percentage of measurement error in ARC map while using any kriging technique it assumes that the measurement error equals to 100% of the error. this negatively ompact the generated geostatistical layer contour lines and the values at the easured rain gauges are significantly impacted. when i use a zero percent error the generated geostatistical layers for both t universal and ordinary kriging are identical
Good question! Dr. Journel told us not to put kriging estimates in maps! They are the best estimates at each location, but jointly they are incorrect, because they do not honor the histogram nor the variogram. The kriging variance is the missing variance in kriging and a measure of uncertainty in the estimate.
Howdy Olatunde, you could cite the book, Pyrcz and Deutsch, 2014, Geostatistical Reservoir Modeling, 2nd edition. I'm glad the content is useful to you, Michael
Great lecture !! I've noticed a small mistake at 14:17 : The indexes are C(u_i,u_j) instead of C(u_i,u_i) on the two equations ! Thanks again I hope I'm not wrong
Howdy Solima, you're welcome and check out my ExcelNumericalDemos repository, I have simple kriging, indicator kriging and collocated cokriging by-hand. Hope this helps.
@@CK-vy2qv, you are correct. That is Darby, my rescue dog! She likes to join in my recorded lectures. I'm glad that you are finding the content useful!
Hello sir. Thanks again for your great and straightforward content on geostatistics. I owe you a debt of gratitude. I had a question about ordinary kriging. In the 2nd half of this video on kriging theory by Luc Anselin, he said that in ordinary kriging the mean is constant and does not vary locally, there exists a stationary condition, and it is the case in a universal kriging model that the mean varies locally, while you said that in ordinary kriging we relax the stationarity condition: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-AoIUcE0vvq8.html Am I right or this is some kind of misunderstanding?
