A spatio-temporal model for Red Sea surface temperature anomalies

Abstract

This paper details the approach of team Lancaster to the 2019 EVA data challenge, dealing with spatio-temporal modelling of Red Sea surface temperature anomalies. We model the marginal distributions and dependence features separately; for the former, we use a combination of Gaussian and generalised Pareto distributions, while the dependence is captured using a localised Gaussian process approach. We also propose a space-time moving estimate of the cumulative distribution function that takes into account spatial variation and temporal trend in the anomalies, to be used in those regions with limited available data. The team's predictions are compared to results obtained via an empirical benchmark. Our approach performs well in terms of the threshold-weighted continuous ranked probability score criterion, chosen by the challenge organiser.

Quantile based modelling of diurnal temperature range with the five-parameter lambda distribution

“…Davison and Huser (2019) model extreme temperatures in Spain with the GPD and find that the tail parameter, which we denote by λ 5 , is approximately equal to 0.4. Castro-Camilo et al (2021) and Rohrbeck et al (2021) model Red Sea temperatures with the GPD and find that the tail parameter is larger than −0.1. O'Sullivan et al ( 2020) model temperature extremes in Dublin and find that the posterior median of the tail parameter is larger than 0.1.…”

“…Castro-Camilo et al, 2021;Davison & Huser, 2019). Aditionally, Rohrbeck et al (2021) model daily temperature in the Red Sea by assuming that both the upper and lower tails of daily temperature can be modelled using the GPD, and Stein (2021) proposes to model climatological phenomena using parametric distributions that behaves like the GPD in both tails, and uses this approach to model daily average temperature near Calgary during winter. The GPD can be described through its quantile function (e.g.…”

Quantile based modeling of diurnal temperature range with the five‐parameter lambda distribution

“…Castro‐Camilo et al, 2021; Davison & Huser, 2019). Additionally, Rohrbeck et al (2021) model daily temperature in the Red Sea by assuming that both the upper and lower tails of daily temperature can be modeled using the generalized Pareto distribution, and Stein (2021b) proposes to model climatological phenomena using parametric distributions that behaves like the generalized Pareto distribution in both tails, and uses this approach to model daily average temperature near Calgary during winter. The generalized Pareto distribution can be described through its quantile function (e.g.…”

“…Davison and Huser (2019) model extreme temperatures in Spain with the generalized Pareto distribution and find that the tail parameter, which we denote by , is approximately equal to 0.4. Castro‐Camilo et al (2021); Rohrbeck et al (2021) model Red Sea temperatures with the generalized Pareto distribution and find that the tail parameter is larger than . O'Sullivan et al (2020) model temperature extremes in Dublin and find that the posterior median of the tail parameter is larger than .…”

“…The generalized Pareto distribution is often used for estimating extremely large or small quantiles, and has been much used for modeling both the upper and lower tails of temperature distributions (e.g. Rohrbeck et al, 2021; Stein, 2021a). Thus, the FPL distribution is a natural choice for modeling the difference between daily maximum and daily minimum temperature if one considers these as extreme upper and lower quantiles in the daily temperature distribution.…”