mixle.analysis.max_stable module¶
Max-stable processes for spatial extremes: the Smith (Gaussian-storm) model.
Block maxima of a spatial field (annual flood peaks, peak seismic amplitude, extreme porosity) are
spatially dependent, and that dependence has its own limit law – a max-stable process – which the
ordinary GEV/GPD (treated independently per site) misses. The Smith model is the canonical one:
Z(s) = max_i xi_i * phi_Sigma(s - U_i) over a Poisson storm process, giving unit-Frechet margins and a
closed-form pairwise dependence. The extremal coefficient theta(h) in [1, 2] summarizes it: 1 = full
dependence (extremes always co-occur), 2 = independence. Part of the earth-science/UQ work (Phase 6).
- class SmithMaxStable(sigma)[source]
Bases:
objectThe Smith max-stable process with Gaussian storm-profile covariance
sigma(d x d, SPD).A spatial process, not an i.i.d. leaf distribution: its full likelihood is intractable, so it is not a
SequenceEncodableProbabilityDistribution– it exposes the things that do have closed forms (extremal_coefficient,bivariate_cdf) plus asampler, and is fitted by the module-levelfit_smith_maxstable()(composite/madogram estimation), mirroring the functional fit style of the other non-leaf spatial models. Margins are unit Frechet; spatial dependence grows withsigma.- Parameters:
sigma (np.ndarray)
- extremal_coefficient(h)[source]
theta(h) = 2 * Phi(a/2)withathe Mahalanobis lag length – 1 at h=0 (full dependence) rising to 2 as the lag grows (independence).
- bivariate_cdf(z1, z2, h)[source]
P(Z(s) <= z1, Z(s+h) <= z2) = exp(-V(z1, z2))– the Smith bivariate distribution.
- fit_smith_maxstable(locations, fields)[source]
Fit an isotropic Smith max-stable process (
sigma = s^2 I) to replicated spatial extremes.locationsis(n_locations, d)andfieldsis(n_replicates, n_locations)of block maxima. Estimation matches the binned empirical extremal coefficient (from the F-madogram) to the model2 Phi(|h| / (2 s)). Returns aSmithMaxStable.