Intraseasonal descriptors and rainfall extremes in austral summer over South Africa: observations and meso-scale modelling

Started in december 2018

Funding: I-SITE Junior Fellowship IMVULA

Supervisors: Benjamin Pohl and Julien Pergaud

Abstract

Rainfall extremes are of major and increasing importance in semi-arid countries and their variability has strong implications for water resource and climate impacts on the local societies and environment. Here, we examine intraseasonal descriptors (ISDs) and wet extremes in austral summer rainfall (November−February) over South Africa (SA). Using daily observations from 225 rain gauges and ERA5 reanalysis between 1979 and 2015, we propose a novel typology of wet extreme events based on their spatial fraction, thus differentiating large- and small-scale extremes. Long-term variability of both types of extreme rainfall events is then extensively discussed in the context of ISDs. Following the definition of a novel typology of rainfall extremes, disentangling large- and small-scale events we further examine the relationship between these two types of rainfall extremes and different modes of climate variability at different timescales. At low-frequencies, rainfall extremes are assessed at interannual (IV: 2−8 years) and quasi-decadal (QDV: 8−13 years) timescales, which are primarily associated with the El Niño Southern Oscillation (ENSO) and the Interdecadal Pacific Oscillation (IPO), respectively. At sub-seasonal timescales, the typology of rainfall extremes is analysed depending on the synoptic configurations, as inferred by seven convective regimes including Tropical Temperate Troughs (TTTs: 3–7 days), and the intraseasonal variability associated with the Madden-Julien Oscillation (MJO: 30–60 days). To identify potentially high-impact rainfall spells, we introduce duration into the definition of extreme rainfall typology. Large-scale longest-lived events are then considered as case studies of potentially high-impact rainfall spells and are selected for meso-scale modelling. To that end, we use state-of-the-art Weather Research and Forecasting (WRF version 4.2.1) model which is widely used for both atmospheric research and operational forecasting applications. ERA5 reanalysis is first used to drive WRF simulations with several experimental setups on different case studies to obtain the finest WRF configuration. The optimal experimental design is then used to drive WRF simulations using coarse resolution global forecast i.e., Re-forecast version 2 (RF2 hereinafter) developed by National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESLR). Here, we attempt to investigate if the properties of such major events can be improved using convection-permitting downscaling of these global grids in terms on intensity and location.

The results demonstrate that using 7 % of spatial fraction simultaneously exceeding the local threshold of the 90th percentile produces remarkable results in characterizing rainfall extremes into large- and small-scale extremes. Austral summer total rainfall is found to be primarily shaped by large-scale extremes which constitute more than half of the rainfall amount under observation, and nearly half in ERA5. Observation (ERA5) shows an average of 8±5 (20±7) days per season associated with large-scale extremes, which are comprised in 5±3 (10±3) spells with an average persistence of at least 2 days. Overall, we find a strong dependence of total rainfall on the number of wet days and wet spells that are associated with large-scale extremes. We also find that large- and small-scale extremes are well-organized and spatially coherent in nature yet extreme conditions during small-scale events are found sporadic over the region, contrasting with large-scale events for which extreme conditions are found over a larger and coherent region. An added value of this work resides in the first presentation of a detailed mapping of rainfall variability over South Africa, including large- and small-scale extreme events, as well as non-extreme rainfall contribution. Such studies have immediate and considerable implications for theoretical and applied climate variability-based studies.

At the IV timescale, the occurrence of large-scale extremes is substantially higher during its wet seasons thereby suggesting a 400 % rise in the occurrence large-scale extremes as compared to its dry seasons. QDV timescale found to be mostly related to modulate small-scale extremes during its wet seasons. Teleconnections with global sea surface temperature (SSTs) confirm that La Niña conditions favour overall wet conditions and extremes in South Africa. The numbers of large-scale extremes are consistently related to warmer SSTs in the North Atlantic, while their link with warmer Indian and tropical South Atlantic oceans is found to be statistically independent of the state of ENSO. At the sub-seasonal timescales, large-scale extremes largely occur during synoptic regimes #3 to #5 whereas small-scale extremes are nearly equiprobable during all regimes. The occurrence of large-scale extremes during continent rooted TTT is further enhanced during the locally wet phases of the MJO and is symmetrically weakened during its dry phases. The extensive characterization of rainfall extremes carried out in reference to low-frequency and sub-seasonal timescales of variability are crucial in promoting long-term multi-year seamless forecasts for the region on one hand, and sub-seasonal forecasts on the other hand.

Extreme rainfall spell during February 2000 season (06 February through 10 February 2000) is then selected as a case study for downscaling using Weather Research and Forecasting Model (WRF). The WRF model is first forced by ERA5 reanalysis and then forced by one control-run and four ensemble members of NOAA’s Re-forecast version 2 (RF2), with a lead time up to 9 days. Concerning using ERA5 as forcings, the location and the intensity shown by raw ERA5 is better in parent domain of WRF and much resolved in the inner domain. The storm center on the first day of the event shown by TRMM dataset is well simulated at 9Km WRF domain while at 3Km domain show substantially better characteristics in terms of rainfall amounts and the correct location. The southward propagation of the storm as shown by raw ERA5 is also well captured by both domains of WRF. RF2 raw field reproduced skilfully the rainfall pattern, thereby suggesting a considerable performance to use it as a forcings to downscale it using WRF. Preliminary results of downscaling of February 2000 event using global grid of RF2 suggest that the forecast of such events can be substantially improved by WRF. The maximum rainfall during the highest intensity day of the event is well resolved by WRF as compared to raw RF2 grids up to 7 to 8 days lead time however with some uncertainties between ensemble members. Further analysis yet to be conducted to evaluate the biases, uncertainties, and associated dynamics.

Jury

Chris Reason, université de Cape Town (Afrique du Sud) – reviewer
Willem Landman, université de Pretoria (Afrique du Sud) – reviewer
Bastien Dieppois, université de Coventry (Royaume-Uni) – examiner
Pierre Camberlin, université de Bourgogne – president
Benjamin Pohl, laboratoire Biogéosciences – supervisor
Julien Pergaud, laboratoire Biogéosciences – cosupervisor

extrait:

lien_externe:

titre:

Descripteurs in-saisonniers de la saison des pluies en Afrique australe : observations et modélisation à moyenne échelle

date_de_debut_these:

décembre 2018

nom:

Ullah

date_de_debut_these_numerique:

201812

kc_data:

a:8:{i:0;s:0:"";s:4:"mode";s:0:"";s:3:"css";s:0:"";s:9:"max_width";s:0:"";s:7:"classes";s:0:"";s:9:"thumbnail";s:0:"";s:9:"collapsed";s:0:"";s:9:"optimized";s:0:"";}

kc_raw_content:

Intraseasonal descriptors and rainfall extremes in austral summer over South Africa: observations and meso-scale modelling

Started in december 2018

Funding: I-SITE Junior Fellowship IMVULA

Supervisors: Benjamin Pohl and Julien Pergaud

 

Abstract

Rainfall extremes are of major and increasing importance in semi-arid countries and their variability has strong implications for water resource and climate impacts on the local societies and environment. Here, we examine intraseasonal descriptors (ISDs) and wet extremes in austral summer rainfall (November−February) over South Africa (SA). Using daily observations from 225 rain gauges and ERA5 reanalysis between 1979 and 2015, we propose a novel typology of wet extreme events based on their spatial fraction, thus differentiating large- and small-scale extremes. Long-term variability of both types of extreme rainfall events is then extensively discussed in the context of ISDs. Following the definition of a novel typology of rainfall extremes, disentangling large- and small-scale events we further examine the relationship between these two types of rainfall extremes and different modes of climate variability at different timescales. At low-frequencies, rainfall extremes are assessed at interannual (IV: 2−8 years) and quasi-decadal (QDV: 8−13 years) timescales, which are primarily associated with the El Niño Southern Oscillation (ENSO) and the Interdecadal Pacific Oscillation (IPO), respectively. At sub-seasonal timescales, the typology of rainfall extremes is analysed depending on the synoptic configurations, as inferred by seven convective regimes including Tropical Temperate Troughs (TTTs: 3–7 days), and the intraseasonal variability associated with the Madden-Julien Oscillation (MJO: 30–60 days). To identify potentially high-impact rainfall spells, we introduce duration into the definition of extreme rainfall typology. Large-scale longest-lived events are then considered as case studies of potentially high-impact rainfall spells and are selected for meso-scale modelling. To that end, we use state-of-the-art Weather Research and Forecasting (WRF version 4.2.1) model which is widely used for both atmospheric research and operational forecasting applications. ERA5 reanalysis is first used to drive WRF simulations with several experimental setups on different case studies to obtain the finest WRF configuration. The optimal experimental design is then used to drive WRF simulations using coarse resolution global forecast i.e., Re-forecast version 2 (RF2 hereinafter) developed by National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESLR). Here, we attempt to investigate if the properties of such major events can be improved using convection-permitting downscaling of these global grids in terms on intensity and location.

The results demonstrate that using 7 % of spatial fraction simultaneously exceeding the local threshold of the 90th percentile produces remarkable results in characterizing rainfall extremes into large- and small-scale extremes. Austral summer total rainfall is found to be primarily shaped by large-scale extremes which constitute more than half of the rainfall amount under observation, and nearly half in ERA5. Observation (ERA5) shows an average of 8±5 (20±7) days per season associated with large-scale extremes, which are comprised in 5±3 (10±3) spells with an average persistence of at least 2 days. Overall, we find a strong dependence of total rainfall on the number of wet days and wet spells that are associated with large-scale extremes. We also find that large- and small-scale extremes are well-organized and spatially coherent in nature yet extreme conditions during small-scale events are found sporadic over the region, contrasting with large-scale events for which extreme conditions are found over a larger and coherent region. An added value of this work resides in the first presentation of a detailed mapping of rainfall variability over South Africa, including large- and small-scale extreme events, as well as non-extreme rainfall contribution. Such studies have immediate and considerable implications for theoretical and applied climate variability-based studies.

At the IV timescale, the occurrence of large-scale extremes is substantially higher during its wet seasons thereby suggesting a 400 % rise in the occurrence large-scale extremes as compared to its dry seasons. QDV timescale found to be mostly related to modulate small-scale extremes during its wet seasons. Teleconnections with global sea surface temperature (SSTs) confirm that La Niña conditions favour overall wet conditions and extremes in South Africa. The numbers of large-scale extremes are consistently related to warmer SSTs in the North Atlantic, while their link with warmer Indian and tropical South Atlantic oceans is found to be statistically independent of the state of ENSO. At the sub-seasonal timescales, large-scale extremes largely occur during synoptic regimes #3 to #5 whereas small-scale extremes are nearly equiprobable during all regimes. The occurrence of large-scale extremes during continent rooted TTT is further enhanced during the locally wet phases of the MJO and is symmetrically weakened during its dry phases. The extensive characterization of rainfall extremes carried out in reference to low-frequency and sub-seasonal timescales of variability are crucial in promoting long-term multi-year seamless forecasts for the region on one hand, and sub-seasonal forecasts on the other hand.

Extreme rainfall spell during February 2000 season (06 February through 10 February 2000) is then selected as a case study for downscaling using Weather Research and Forecasting Model (WRF). The WRF model is first forced by ERA5 reanalysis and then forced by one control-run and four ensemble members of NOAA’s Re-forecast version 2 (RF2), with a lead time up to 9 days. Concerning using ERA5 as forcings, the location and the intensity shown by raw ERA5 is better in parent domain of WRF and much resolved in the inner domain. The storm center on the first day of the event shown by TRMM dataset is well simulated at 9Km WRF domain while at 3Km domain show substantially better characteristics in terms of rainfall amounts and the correct location. The southward propagation of the storm as shown by raw ERA5 is also well captured by both domains of WRF. RF2 raw field reproduced skilfully the rainfall pattern, thereby suggesting a considerable performance to use it as a forcings to downscale it using WRF. Preliminary results of downscaling of February 2000 event using global grid of RF2 suggest that the forecast of such events can be substantially improved by WRF. The maximum rainfall during the highest intensity day of the event is well resolved by WRF as compared to raw RF2 grids up to 7 to 8 days lead time however with some uncertainties between ensemble members. Further analysis yet to be conducted to evaluate the biases, uncertainties, and associated dynamics.

 

Jury

Chris Reason, université de Cape Town (Afrique du Sud) - reviewer
Willem Landman, université de Pretoria (Afrique du Sud) - reviewer
Bastien Dieppois, université de Coventry (Royaume-Uni) - examiner
Pierre Camberlin, université de Bourgogne - president
Benjamin Pohl, laboratoire Biogéosciences - supervisor
Julien Pergaud, laboratoire Biogéosciences - cosupervisor