Short course on space-time modeling

23
Short course on space-time modeling Instructors: Peter Guttorp Johan Lindström Paul Sampson

description

Short course on space-time modeling. Instructors: Peter Guttorp Johan Lindström Paul Sampson. Schedule. 9:10 – 9:50 Lecture 1: Kriging 9:50 – 10:30 Lab 1 10:30 – 11:00 Coffee break 11:00 – 11:45 Lecture 2: Nonstationary covariances 11:45 – 12:30 Lecture 3: Gaussian Markov random fields - PowerPoint PPT Presentation

Transcript of Short course on space-time modeling

Page 1: Short course on  space-time modeling

Short course on space-time modeling

Instructors:Peter GuttorpJohan LindströmPaul Sampson

Page 2: Short course on  space-time modeling

Schedule9:10 – 9:50 Lecture 1: Kriging9:50 – 10:30 Lab 110:30 – 11:00 Coffee break11:00 – 11:45 Lecture 2:

Nonstationary covariances11:45 – 12:30 Lecture 3: Gaussian

Markov random fields12:30 – 13:30 Lunch break13:30 – 14:20 Lab 214:20 – 15:05 Lecture 4: Space-

time modeling15:05 – 15:30 Lecture 5: A case

study15:30 – 15:45 Coffee break15:45 – 16:45 Lab 3

Page 3: Short course on  space-time modeling

Kriging

Page 4: Short course on  space-time modeling

The geostatistical model

Gaussian processμ(s)=EZ(s) Var Z(s) < ∞Z is strictly stationary if

Z is weakly stationary if

Z is isotropic if weakly stationary and

Page 5: Short course on  space-time modeling

The problem

Given observations at n locationsZ(s1),...,Z(sn)

estimateZ(s0) (the process at an unobserved

location)

(an average of the process)

In the environmental context often time series of observations at the locations.

or

Page 6: Short course on  space-time modeling

Some history

Regression (Bravais, Galton, Bartlett)Mining engineers (Krige 1951, Matheron, 60s)Spatial models (Whittle, 1954)Forestry (Matérn, 1960)Objective analysis (Gandin, 1961)More recent work Cressie (1993), Stein (1999)

Page 7: Short course on  space-time modeling

A Gaussian formula

If

then

Page 8: Short course on  space-time modeling

Simple krigingLet X = (Z(s1),...,Z(sn))T, Y = Z(s0), so that

μX=μ1n, μY=μ, ΣXX=[C(si-sj)], ΣYY=C(0), and

ΣYX=[C(si-s0)].

Then

This is the best unbiased linear predictor when μ and C are known (simple kriging).

The prediction variance is

Page 9: Short course on  space-time modeling

Some variants

Ordinary kriging (unknown μ)

where

Universal kriging (μ(s)=A(s)βfor some spatial variable A)

where Still optimal for known C.

Page 10: Short course on  space-time modeling

Universal kriging variance

simple kriging variance

variability due to estimating μ

Page 11: Short course on  space-time modeling

The (semi)variogram

Intrinsic stationarityWeaker assumption (C(0) needs not exist)Kriging predictions can be expressed in terms of the variogram instead of the covariance.

Page 12: Short course on  space-time modeling

The exponential variogram

A commonly used variogram function is γ(h) = σ2 (1 – e–h/ϕ. Corresponds to a Gaussian process with continuous but not differentiable sample paths.More generally,

has a nugget τ2, corresponding to measurement error and spatial correlation at small distances.

Page 13: Short course on  space-time modeling

Nugget Effective range

Sill

Page 14: Short course on  space-time modeling

Ordinary kriging

where

and kriging variance

Page 15: Short course on  space-time modeling

An example

Precipitation data from Parana state in Brazil (May-June, averaged over years)

Page 16: Short course on  space-time modeling

Variogram plots

Page 17: Short course on  space-time modeling

Kriging surface

Page 18: Short course on  space-time modeling

Bayesian kriging

Instead of estimating the parameters, we put a prior distribution on them, and update the distribution using the data.Model:

Matrix withi,j-elementC(si-sj; φ)(correlation)

measurementerror

θβσφτT

(Z(s1)...Z(sn))T

Page 19: Short course on  space-time modeling

Prior/posterior of φ

Page 20: Short course on  space-time modeling

Estimated variogram

ml

Bayes

Page 21: Short course on  space-time modeling

Prediction sites

1

2

3

4

Page 22: Short course on  space-time modeling

Predictive distribution

Page 23: Short course on  space-time modeling

References

A. Gelfand, P. Diggle, M. Fuentes and P. Guttorp, eds. (2010): Handbook of Spatial Statistics. Section 2, Continuous Spatial Variation. Chapman & Hall/CRC Press.

P.J. Diggle and Paulo Justiniano Ribeiro (2010): Model-based Geostatistics. Springer.