Dependence Types in a Binarized Precipitation Network. Issue 23 (26th November 2020)

• Record Type: Journal Article
• Authors: De Michele, C.; Meroni, V.; Rahimi, L.; Deidda, C.; Ghezzi, A.
• Is Part Of: Geophysical research letters. Volume 47:Issue 23(2020)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-11-26
• Subjects: compound events -- dependence -- precipitation network -- binary variables; Geophysics -- Periodicals; Planets -- Periodicals; Lunar geology -- Periodicals; 550
• DOI: 10.1029/2020GL090196
• Languages: English
• ISSNs: 0094-8276

Abstract

Abstract: In a network of binarized precipitation (i.e., wet or dry value), the connection or dependence between each pair of nodes can occur following one or more of the following conditions: wet‐wet, dry‐dry, wet‐dry, or dry‐wet. Here, we firstly investigate the different types of dependence, year by year, within a precipitation network of binarized variables. We compare the sample estimate of the probability of co‐occurrence (or occurrence with a lag time within ±3 days) of each of the four possible combinations with respect to the correspondent confidence interval in hypothesis of independence. We develop a procedure to efficiently assess the dependence behavior of all couples of nodes within the network and apply the methodology to a network of rain gauges covering Europe and north Africa. Plain Language Summary: Meteorological connections are important issues of the climate system because they can help identifying geographic areas where compound weather events could occur. These are natural hazards characterized by extreme and devastating impacts. The typology of compound events (e.g., the co‐occurrence, or subsequence, of two floods, or two droughts, or a flood and a drought, in two geographic areas, among others) is a function of the type of dependence between the variables involved. Here, we investigate the different types of pairwise dependence, year by year, within a precipitation network with respect to wet‐wet, dry‐dry, wet‐dry, and dry‐wet co‐occurrences (or occurrences with a lag time within ±3 days), introducing a general methodology based on probability measures. This helps identifying situations which could lead respectively to compound floods or compound droughts, as well as connections between floods and droughts. We apply the methodology to a network of rain gauges covering Europe and North Africa. In order to optimize the computational times, we develop also a procedure to efficiently calculate the significance of the dependence between each pair of nodes within the network. The results show how it is possible to identify geographic areas where the wet‐wet and/or dry‐dry occurrences are frequent, so where we should expect compound floods and/or compound drought/low flow conditions, and where wet‐dry and/or dry‐wet conditions occur: mainly wet conditions in south and dry conditions in north. Key Points: We investigate the wet‐wet, dry‐dry, and wet‐dry (dry‐wet) pairwise dependence within a precipitation network We develop a methodology and a procedure to test the dependence between each pair of nodes and applied it to a network of rain gauges covering Europe and North Africa We identify geographic areas where the wet‐wet and/or dry‐dry connections are frequent and where the wet‐dry and/or dry‐wet connections are observed