Atmospheric Rivers in Africa Observed with GNSS-RO and Reanalysis Data

Abstract

Atmospheric Rivers (ARs) transport significant amounts of moisture and cause extreme precipitation events, yet their behavior over Africa is not well understood. This study addresses this gap by analyzing the occurrence, seasonal variability, and spatial dynamics of ARs across the continent from 2009 to 2019. Utilizing ERA5 reanalysis data, Global Navigation Satellite Systems Radio Occultation (GNSS RO) measurements, and the Image-Processing-based Atmospheric River Tracking (IPART) method, distinct seasonal AR patterns are identified. Southern Africa experiences peak activity during austral summer, while AR occurrence in Northern Africa peaks in boreal winter and spring, aligning with regional rainy seasons. Moisture sources include the Atlantic Ocean, the Arabian Sea, and the Red Sea. A comparison of ERA5 Integrated Water Vapor (IWV) estimates with high-resolution GNSS RO data shows that both datasets effectively capture broad-scale moisture patterns. However, ERA5 consistently delivers higher IWV values compared to GNSS RO, which is likely due to underrepresentation of GNSS RO IWV values, since profiles generally do not reach all the way down to the surface—but also due to an overrepresentation of humidity in the ERA5 reanalyses. Understanding AR dynamics in Africa is essential to improve climate resilience, water management and understanding extreme precipitation events.

1. Introduction

Atmospheric Rivers (ARs) are long, narrow corridors of concentrated water vapor that transport moisture over long distances. They are defined by a maximum width of 500 km and a minimum length of 2000 km in the (lower) troposphere. Typically, ARs are defined by thresholds; for Integrated Vapor Transport (IVT), these values need to be above 250 kg m⁻¹ s⁻¹. Each AR can carry water vapor quantities comparable to the flow of the Mississippi River, with three to five events typically present per hemisphere at any time [1]. Known for influencing precipitation, ARs are responsible for heavy rainfall and flooding, particularly in mid-latitude and subtropical regions [2]. While AR dynamics are well studied in America and Europe, studies in Africa remain limited. Globally, ARs are crucial in the hydrological cycle, contributing to water availability and extreme weather events. Their ability to transport large volumes of water vapor from tropical oceanic sources to inland areas has broad implications for water resource management, forecasting extreme weather and recognizing climate impacts [3]. Understanding their role in such processes underscores the importance of an in-depth understanding of AR characteristics specific to Africa, where weather patterns are uniquely complex.

1.1. Atmospheric Rivers over Africa—Moisture Sources

In Africa, ARs influence rainfall patterns but have primarily been studied on a case-by-case basis, with limited research examining their behavior over an extended period across the entire continent. Key moisture sources for African ARs include the North and South Atlantic Ocean, the Arabian Sea, and the Red Sea [4,5]. The North Atlantic Ocean plays a particularly important role in AR formation, as water vapor is transported eastward towards Africa, influencing rainfall in regions such as the Middle East and Northern Africa (MENA). In Southern Africa, ARs typically draw moisture from the South Atlantic Ocean and tropical areas, interacting with extratropical cyclones and cold fronts to produce heavy rainfall.

The general development includes a source of water vapor, like an ocean or sea, providing essential moisture, while large-scale low-pressure systems, such as cyclones, help organize this moisture into narrow vapor corridors. Wind shear, often generated by jet streams, maintains the elongated structure of ARs and drives rapid moisture transport [6].

For Northern Africa, Egypt and the Middle East, the North Atlantic plays a particularly important role in AR formation, as moisture-rich air is transported eastward, impacting rainfall in regions like Mauritania and Egypt [4]. ARs in the MENA region form primarily in the lower troposphere, where tropical moisture is transported northward by low-level southwesterly winds, often enhanced by mid-level anticyclonic circulations over the Arabian Sea [7]. These circulations channel moisture from the Gulf of Aden and the Red Sea. Additionally, the interaction of upper-level cyclonic systems with mid-level anticyclonic patterns supports moisture convergence, while the merging of the subtropical and polar jet streams over Africa strengthens meridional flow, facilitating inland moisture transport and AR development [5].

In Northern Africa, ARs are influenced by the North Atlantic Oscillation (NAO). A positive NAO phase strengthens westerlies, pushing ARs farther inland, reaching Mauritania, Senegal, and the Middle East, with the subtropical and polar jet streams enhancing moisture transport [4]. The polar jet stream, which is typically associated with mid- and high-latitude regions, can occasionally extend over far northern Africa, particularly during the boreal winter months. This occurs when the jet stream dips southward, bringing with it strong upper-level winds that can influence weather patterns in the region. While the polar jet stream is more commonly associated with higher latitudes, its occasional presence over northern Africa highlights the dynamic nature of atmospheric circulation and its impact on regional climate.

In Southern Africa, ARs draw moisture from the South Atlantic and tropical sources, interacting with extratropical cyclones and cold fronts to deliver substantial rainfall, especially during boreal winter [8]. The analysis by Ramos et al. (2019) [8] identifies four main moisture sources: (1) the western South Atlantic near Brazil, where tropical convergence enhances moisture uptake; (2) the eastern South Atlantic near the Cape Agulhas, linked to the Agulhas Current retroflection; (3) the Agulhas Current itself, which supplies a steady moisture stream along South Africa’s east coast; and (4) continental sources in northern and northwestern South Africa, Namibia, and The Republic of Botswana. This moisture transport is further intensified by the South American Low-Level Jet (SALLJ), which channels Amazonian moisture to the South Atlantic, reinforcing AR-driven rainfall in Southern Africa. The pathways are shaped by the South Atlantic Subtropical High (SASH) and interactions with extratropical cyclones and cold fronts [7]. These ARs move along a southwest-to-northeast path, drawing moisture from the South Atlantic and occasionally from South America. In addition, mountain ranges intensify AR-driven precipitation through orographic lift, like the Cape Fold Mountains [8].

There, they contribute to winter rainfall, since ARs are most common in early austral winter (May to September) [7,9,10]. For instance, Blamey et al. (2018) [7] observed that Atmospheric Rivers were responsible for approximately 70% of the 50 most extreme winter rainfall events, emphasizing their role in contributing to heavy rainfall and flooding risks.

1.2. Seasonal and Interannual Variability

ARs in both Northern and Southern Africa are subject to seasonal and interannual variability, largely driven by the interactions between global climate oscillations and regional weather patterns. According to the available literature, AR activity in Northern Africa peaks during boreal fall and winter, when the Azores High retreats, allowing more moisture-laden air to penetrate the region [4]. Conversely, during the boreal summer, AR activity decreases due to the dominance of the Azores High and stable atmospheric conditions.

In Southern Africa, AR activity is highly seasonal, peaking during the austral winter (May to September), with extratropical cyclones and cold fronts driving moisture from the South Atlantic toward the southwestern coast of South Africa [8]. The interannual variability of ARs is further influenced by large-scale climate patterns, including El Niño-Southern Oscillation (ENSO) and the Southern Annular Mode (SAM), which affect both the frequency and intensity of ARs across the African continent [11].

1.3. Global Navigation Satellite Systems Radio Occultation (GNSS RO)

GNSS RO (abbreviated as “RO” from now on) is a satellite-based remote sensing technique that provides high-vertical-resolution profiles of atmospheric refractivity, temperature, pressure, and water vapor by measuring the bending of GNSS signals as they pass through the atmosphere [12]. RO has been widely utilized in atmospheric studies due to its global coverage, all-weather capability, and long-term stability, making it an essential dataset for climate research and numerical weather prediction (NWP). RO-derived temperatures are highly accurate and consistent in the upper troposphere and lower stratosphere [13,14], while the quality of RO-derived water vapor is highest in the lower and middle troposphere [15,16].

However, RO measurements can underestimate Integrated Water Vapor (IWV) due to missing data in the lowest few hundred meters of the troposphere, when the radio signals do not reach all the way down to the surface [17,18,19,20,21,22]. Therefore, RO data tend to underestimate IWV compared to satellite-based instruments like the Special Sensor Microwave Imager/Sounder (SSMI/S), but still closely align closely (~95%) with collocated ECMWF data [23].

In this study, we used RO data from multiple satellite missions to compare them with the representation of moisture transport based on ERA5 reanalysis data. RO water vapor retrievals commonly use a 1-dimensional variational (1DVAR) assimilation technique, where RO refractivity profiles are combined with a background to generate water vapor profiles [23]. While this method ensures consistency, it also introduces the possibility of biases propagating from the background field, which must be carefully considered when comparing RO with ERA5 IWV, even more if ERA5 data have been used (as in our study) as a priori. However, even though our RO data are not independent from ERA5, we know from previous studies [23] that 1DVAR retrievals from two different centers agree surprisingly well, even if the a priori humidity profiles are totally different—particularly in the altitude range, where the retrieval has a high “trust” in the measured data, expressed as a low assumed error for the measured data within the 1DVAR scheme (typically in the altitude range between about 0.5 km and about 7 km).

1.4. AR Tracking

Tracking ARs is crucial for understanding moisture transport and its role in extreme weather events. Various AR detection methods have been developed, with most relying on Integrated Vapor Transport (IVT) and IWV thresholds [24]. However, these methods often struggle with high spatial variability and false detections due to noisy or discontinuous data.

The Image-Processing-based Atmospheric River Tracking (IPART) method, developed by Xu et al. (2020) [25], enhances AR detection by focusing on the spatial and temporal characteristics of ARs. IPART utilizes the Top-Heat by Reconstruction (THR) algorithm, which identifies moisture structures even in noisy data, highlighting regions of high moisture continuity. This method constructs a topological graph of the AR’s moisture flow, accurately tracking the AR’s central path or “axis” as it progresses [24]. Unlike traditional threshold-based methods, IPART incorporates the Top-Heat by Reconstruction (THR) algorithm, which applies image-processing techniques such as grayscale erosion and dilation to identify high-moisture structures even in fragmented data; offers topological graph mapping, which tracks AR central axes and moisture flow dynamically [24]; and has improved tracking accuracy over land, making it well suited for analyzing AR behavior in Africa.

Given its robustness, we chose IPART as the primary AR detection method, ensuring reliable tracking of AR events and their associated moisture transport over Africa [25]. Additionally, the ARTracks catalogue, which integrates IPART with ERA5, was used to identify AR landfall locations and validate seasonal patterns.

1.5. Societal Relevance

Water resource variability is a critical concern in Africa, especially for agriculture. With the continent’s growing population, climate-driven water variability poses a significant risk. Understanding patterns that lead to flooding is vital for planning climate resilience, especially in regions prone to extreme weather events [26]. With the increasing intensity and frequency of these events, integrating AR dynamics into climate adaptation strategies will be crucial for safeguarding communities and infrastructure [27,28]. In regions heavily reliant on seasonal rainfall, like sub-Saharan Africa, such forecasting tools employing satellite data are essential for climate adaptation and risk management. Enhanced atmospheric and climate research will support more effective public health interventions, sustainable agriculture, and resilient infrastructure, reinforcing the continent’s capacity to respond to climate challenges. ARs contribute to extreme weather events, including floods, which impact both water availability and health, but they also influence dust transport, impacting air quality and health in downstream regions [29,30].

1.6. Research Gaps and Objectives

ARs are gaining increasing recognition as drivers of weather events in Africa, particularly during the winter seasons, contributing to extreme rainfall and flooding [31,32,33,34]. However, despite the growing recognition of their significance in other regions such as North America and Europe, the behavior and characteristics of ARs over Africa remain poorly understood. A major knowledge gap exists when it comes to characterizing the spatiotemporal variability of ARs across the continent and evaluating the accuracy of global reanalysis datasets, such as ERA5, in capturing the complex moisture transport processes associated with these events. This gap is particularly critical given Africa’s vulnerability to extreme weather, making an improved understanding of AR dynamics essential for climate prediction and water resource management.

Our study addresses this gap by investigating AR events over Africa from 2009 to 2019. Utilizing the IPART method, this research seeks to identify AR patterns and assess seasonal and interannual variability across Northern and Southern Africa.

To check the robustness, we compared ERA5 IWV measurements against RO data. These data were chosen because of their global coverage, high vertical resolution and stability over time, together with an approximately uniform distribution, which is important in regions with sparse conventional data, like our study regions. RO provides consistent atmospheric profiles unaffected by clouds or precipitation [23,35]. This comparison aims to study the reliability of ERA5 in capturing moisture levels associated with ARs. By systematically evaluating differences between ERA5 and RO IWV values, we aim to assess the reliability of ERA5 in capturing AR-related moisture transport. This validation may help to improve the representation of AR-driven hydrological processes in climate models and enhancing regional forecasting capabilities.

2. Data and Methods

2.1. Data

The ERA5 reanalysis dataset, providing detailed historical climate data, supports hourly tracking of atmospheric parameters. Complementing this, RO data offer vertical moisture profiles, which are essential for understanding distinctive layers.

The RO technique provides high-vertical-resolution profiles of atmospheric refractivity, temperature, and water vapor, making it a valuable complement to reanalysis datasets and other observation types. This high vertical resolution allows for detailed sampling of the atmosphere, particularly in regions where traditional observations may be sparse or limited.

We use IPART, an image-processing-based technique, to refine AR tracking over land, and the ARtracks catalogue aids in identifying precise AR landfall locations.

The study period of 2009–2019 was selected, as it provides a widespread timeframe to assess AR dynamics over Africa, together with a good coverage of RO data.

2.1.1. ERA5 Reanalysis

The 5th Generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA5) dataset from the Copernicus Climate Change Service (C3S) was used for interpolating and comparing with RO data. Data were drawn from the ERA5 hourly dataset available through the Climate Data Store (CDS), specifically focusing on Total Column Water Vapor (TCWV) to capture AR landfall events (https://www.ecmwf.int/en/forecasts/dataset/ecmwf-reanalysis-v5, last access: 10 March 2024) [36]. With a 0.25° spatial resolution and hourly data, ERA5 provides high-quality historical records dating back to 1940 [37].

2.1.2. GNSS Radio Occultation Data

Reprocessed Level 2 RO data from multiple satellites were used, including TerraSar-X (TXS), Gravity Recovery and Climate Experiment (GRACE), Constellation Observing System for Meteorology, Ionosphere, and Climate-1 (COSMIC-1, 6 satellites), Meteorological Operational Satellites (Metop series), the PAZ (Spanish for “peace”) satellite and the Korean Multi-Purpose Satellite-5 (Kompsat 5). We obtained the data through the COSMIC Data Analysis and Archive Center (CDAAC) (https://data.cosmic.ucar.edu/gnss-ro/, last access: 10 March 2024). These data provide profiles of temperature, pressure, and humidity, enabling high-resolution atmospheric profiling, particularly in the lower atmosphere (~0.1 km resolution near the surface) [38].

The RO technology measures atmospheric refractivity by detecting the bending of GNSS signals as they pass through the atmosphere [13]. Vertical profiles derived from these measurements have been widely used for monitoring atmospheric temperature and water vapor [12,14,15,16]. RO data are often employed to observe ARs and improve AR forecasts, as they provide detailed vertical moisture profiles, global coverage, and all-weather capability [17,18,19,20,21,22,23,24]. The dataset used in our study thus provides detailed vertical moisture profiles, global coverage and all-weather capability crucial for tracking atmospheric changes [24].

The RO water vapor retrievals used in this study are derived from RO refractivity profiles using a 1DVAR technique (see Section 1.3). This method combines the RO observations with a background field, typically provided by gridded numerical weather prediction (NWP) datasets or reanalyses, to iteratively solve for the optimal water vapor profile. The background field serves as an initial guess, which is then adjusted based on the RO refractivity measurements to produce the final water vapor profile.

The CDAAC RO dataset used in this study employs ERA5 data as the background field for 1DVAR. It is important to note that while RO provides valuable independent observations of atmospheric refractivity, the derived water vapor profiles are not “pure observations” but are influenced by the background. This means that the accuracy of the water vapor retrievals depends on both the quality of the RO data and the accuracy of the ERA5 background [23].

The 1DVAR process balances the contributions of the background and the RO observations, where both are inversely weighted with the respective expected errors. The relative contributions of the RO data and the background can be quantified with measures like the “Retrieval to A priori Error Ratio” (RAER) [23].

Given that ERA5 is used as the background for the RO retrievals, some caution is warranted when interpreting comparisons between ERA5 and RO-derived water vapor profiles. Any biases present in the ERA5 background could partly propagate into the RO retrievals, potentially affecting the significance of the comparisons. However, the RO data remain a valuable source of atmospheric information, particularly due to their high vertical resolution and global coverage, which complement reanalysis datasets like ERA5.

2.1.3. Image Processing-Based Atmospheric River Tracking

The IPART method (https://github.com/ihesp/IPART, last access: 14 November 2024). The ARtracks Catalogue, a global resource combining ERA5 reanalysis data with IPART, was used to locate AR landfall points. While IPART provides the foundational methodology for identifying ARs, ARtracks builds on this framework to support the detection, visualization, and tracking of AR events on a global scale. The catalogue provides a detailed AR axis path and landfall location based on IVT and other meteorological data [25]. This catalogue helps with precise analyses of AR impact patterns and their geographic extent (https://github.com/dominiktraxl/artracks, last access: 14 November 2024).

2.2. Data Preprocessing and Quality Control

To ensure accuracy and consistency, data preparation focused on two objectives: (1) statistical analysis of AR occurrences and (2) interpolation and comparison of moisture data between RO observations and ERA5 reanalysis.

ARtracks data (2009–2019) were processed to capture the spatial and temporal characteristics of AR landfalls. Preprocessing included formatting timestamps, filtering by geographic boundaries, and categorizing events into Northern and Southern Africa. Events were further organized by year and month to enable seasonal and regional comparisons. Landfall points were verified using Python 3.12.3. and the Shapely library (v2.0.2), ensuring spatial accuracy and relevance to Africa.

Moisture data from RO and ERA5 were compared to validate moisture transport estimates. ERA5 data were formatted uniformly (latitude, longitude, time, and TCWV), while RO data were cleaned to ensure valid IWV values and remove incomplete records. Nearest-neighbor interpolation aligned ERA5 data with RO observations, retaining only points within a 2.5° spatial and 3 h temporal range. Both datasets were checked for matching units and reviewed for outliers.

2.3. Methodology

The methodology for analyzing AR occurrences over Africa and validating ERA5 reanalysis data with RO observations is outlined here. First, the ARtracks catalogue, combining ERA5 and IPART, is applied to detect, visualize, and statistically analyze AR frequencies and patterns, with a focus on seasonal and regional variations across Africa. Second, a comparative analysis between ERA5 and RO data assesses the accuracy of the reanalysis data in capturing moisture transport, utilizing interpolation and statistical metrics to quantify deviations.

2.3.1. Statistical Analysis of AR Occurrence over Africa

The IPART method was used to analyze IVT anomalies associated with ARs over Africa. Seasonal IVT averages from ERA5 were calculated to identify elevated moisture.

ARtracks data (2009–2019) provided AR occurrences categorized into Northern and Southern Africa. Visualizations such as bar charts and heatmaps illustrated annual, monthly, and seasonal patterns.

AR events with axes extending toward Africa were prioritized, ensuring relevance to the study region. Custom geographic boundaries were applied using Python tools, while data for the Arabian Peninsula were excluded.

The ARtracks catalogue includes each AR event’s date, duration, landfall location, average IVT value, and IVT-weighted centroid coordinates, which are useful for tracking AR movement. Centroids, calculated from IVT vectors (combining wind and specific humidity), represent the AR’s central moisture transport path. They represent the central location of the water vapor transport within the AR. IVT is measured in kg m⁻¹ s⁻¹ and it quantifies the amount of water vapor moving through the atmosphere over a certain distance each second. High IVT values indicate strong moisture transport linked to heavy precipitation.

The IVT is calculated as follows:

$$\mathrm{I}\mathrm{V}\mathrm{T}= \sqrt{{\left({\displaystyle \frac{1}{g}}{\int }_{p_s}^{p_t}qu dp\right)}^2+{\left({\displaystyle \frac{1}{g}}{\int }_{p_s}^{p_t}qv dp\right)}^2}$$

(1)

where q is the specific humidity, u and v are the zonal and meridional wind components, p_s and p_t are surface and top-of-atmosphere pressures, and g is the acceleration due to gravity. This vertical integration captures the total atmospheric moisture transport associated with ARs.

2.3.2. Comparative Analysis Using RO and ERA5 Reanalysis Data

The validation of ERA5 reanalyses with RO observations is divided into two parts: (1) interpolation of RO data to create continuous vertical profiles of atmospheric moisture and (2) direct comparison of these interpolated RO profiles with ERA5 data, aligned in space and time.

Inverse Distance Weighting (IDW) interpolation was used to interpolate ERA5 IWV data to the locations of RO observation points, estimating values at unsampled locations by averaging nearby points weighted by proximity. The nearest neighbors were identified using a k-dimensional tree (KDTree), ensuring efficient retrieval of known values for interpolation. The IDW interpolation was applied using the following formula:

$$\left(\mathrm{s}\right) = {\sum }_{i=1}^n{w}_{i}z\left(s_i\right) = {\displaystyle \frac{\frac{{\sum }_{i=1}^{n}z\left(s_i\right)}{{\left|s-s_i\right|}^P}}{{\sum }_{j=1}^m\frac{1}{{\left|s-s_j\right|}^P}}}$$

(2)

where $s$ is the unsampled location (the RO event location), $s_i$ are the $n$ nearest known data points selected for interpolation. $s_j$ represents the $m$ available known points within the defined search radius (all potential points considered before selecting $n$ neighbors). $z\left(s_i\right)$ is the IWV value at each known point $s_i$. $s-s_i$ represents the distance between the unknown point $s$ and each known point $s_i$. $P$ is the inverse distance weighting exponent, which is set to 2 to balance the influence of nearby and distant points. $m_j$ is the number of ERA5 grid points considered for interpolation. Here, $m_j=4$, since the RO event falls within a $0.25°\times 0.25°$ ERA5 grid cell. The weight function is defined as follows:

$$w= {\displaystyle \frac{1}{{\left|s-s_x\right|}^P}}$$

(3)

To ensure accurate interpolation, a maximum distance tolerance of $2.5°$ was applied, meaning that only known points within this radius were considered for interpolation (i.e., these points form the set of $m$ available known values). The ERA5 dataset has a $0.25°$ spatial resolution, meaning that, typically, four nearest neighbors ($n=4)$ were selected from the set of $m$ available points. However, if fewer than four points were available within the $2.5°$ radius, interpolation was performed using the available points within that range. By setting $P=2,$ the IDW method ensures that closer points contribute significantly more to the interpolated value than those farther away, producing a realistic spatial distribution of IWV in climate data [39,40,41,42,43,44].

For interpolation, RO data from multiple satellite sources are used to calculate IWV by integrating temperature, vapor pressure, and pressure data from wet profiles on the day of the AR event. IWV is calculated as follows:

$$\mathrm{I}\mathrm{W}\mathrm{V} = {\displaystyle \frac{1}{g}}{\int }_{p_s}^{p_t}q dp$$

(4)

The integral (Equation (4)) represents a continuous atmospheric column, while the sum (Equation (5)) approximates this for discrete pressure layers, where $q_{\mathrm{i}}$ is the specific humidity at the ith level and $\Delta p_i$ is the layer thickness [45,46].

$$\mathrm{I}\mathrm{W}\mathrm{V} = {\displaystyle \frac{1}{g}}{\sum }_{i=1}^n{q}_i\Delta p_i$$

(5)

Specific humidity, however, is not directly provided in the CDAAC wet profiles. It was therefore calculated based on [23]. The equation for the specific humidity is given in Equations (6) and (7).

$$q= {\displaystyle \frac{\epsilon · p_v}{p-p_v ·(1-\epsilon )}}$$

(6)

$$\epsilon = {\displaystyle \frac{M_w}{M_d}}$$

(7)

The constant $(\epsilon =0.622)$ represents the ratio of the molar mass of water vapor ($M_w=18.015$ g mol⁻¹) to the molar mass of dry air $(M_d=$28.965 g kg⁻¹) [47]. Profiles of vapor pressure $p_v$ and total air pressure $p$ are taken from the CDAAC wet profile data.

In the second part of the analysis, ERA5 reanalysis data were compared with RO satellite measurements (TXS, GRACE, COSMIC-1, Metop series, PAZ, and Kompsat-5) by interpolating ERA5 IWV values both spatially and temporally to align with RO observation points. Spatial interpolation used a KDTree to identify the four nearest ERA5 grid points within a 2.5° radius, followed by IDW for interpolation to the exact RO event location. Temporal interpolation matched RO observations with the two closest ERA5 timestamps within a ±1.5 h window, using linear interpolation. This approach allowed a direct comparison between interpolated ERA5 IWV and observed RO IWV.

The datasets were then assessed using Mean Bias and Root Mean Square Error (RMSE) to quantify differences in performance, with RMSE values categorized as low (<10%), medium (10–30%), and high (>30%) relative to observed values, based on literature standards [48,49,50], where higher RMSE can be due to differences in atmospheric, spatial, or temporal factors.

2.4. Selected AR Events

We selected AR events spanning both the Northern and Southern Hemispheres from 2009 to 2019, covering austral spring and autumn as well as boreal spring and winter. The events were chosen for their geographic and seasonal diversity. Each event is documented in prior literature, confirming its classification as an AR.

Table 1. Investigated AR events from 2009 to 2019.

Event Name	Date	Affected Region	Area	RO Satellites	Study Domain (Lat°/Lon°)
South Africa 2009	26 September 2009	West coast of South Africa	Southern Africa	Cosmic-1 Metop-A GRACE TSX	−10 to −50/−40 to 30
MENA 2010	15 March 2010	MENA Region	Northern Africa	Cosmic-1 Metop-A TSX	45 to 10/0 to 60
Morocco 2010	30 November 2010	Morocco	Northern Africa	Cosmic-1 Metop-A GRACE TSX	45 to 10/−45 to 15
South Africa 2013	26 May 2013	West coast of South Africa	Southern Africa	Cosmic-1 Metop-A Metop-B GRACE TSX	−5 to −45/−40 to 30
MENA 2017	14 April 2017	Middle East/Iran	Northern Africa	Cosmic-1 Metop-A Metop-B Kompsat5	50 to 10/10 to 60
Mauritania 2019	24 March 2019	Middle East	North Africa	Cosmic-1 Metop-A Metop-B TSX Kompsat5 PAZ	40 to 10/−30 to 60

includes each event’s date that was chosen based on precipitation, affected region, study domain, and satellite sources used for RO observations.

These events highlight key AR dynamics such as moisture uptake, long-distance transport, and interactions with geographic features that intensify precipitation impacts. The selected events, covering diverse regions and seasons, form a robust foundation for analyzing AR behavior across Africa. The study domains for each event are shown in Figure 1.

The South African 2009 event affecting the west coast of South Africa in September 2009 occurred during austral spring. It showed unusual moisture uptake from regions typically outside Southern Africa’s moisture sources, exemplifying teleconnections between South America and South Africa. The event notably impacted the Western Cape Province, which is especially vulnerable to ARs due to its closeness to the South Atlantic Ocean [23].

Taking place in March 2010, in the Middle East and North Africa (MENA) region, the MENA 2010 event occurred in boreal spring. It influenced dust transport and interacted with snowmelt processes in the Near East highlands. The AR primarily drew moisture from the Red Sea and northeastern Africa, impacting the highlands of the Near East and making it a key event for studying AR influences during the snowmelt season [51].

In November 2010, an AR event brought intense rainfall to Morocco, causing substantial flooding, especially in urban areas like Casablanca. The Morocco 2010 event produced precipitation levels nearing 180 mm at specific rain gauges, severely affecting infrastructure. Occurring in late boreal autumn, this event provides insight into North Africa’s pre-winter climate conditions [31].

In the boreal winter of 2011, the Mauritania 2011 event impacted East Sahara, Mauritania, Morocco, and Guinea. The AR demonstrates high frequency and extensive reach. ARs in this area are often influenced by upper-level jet streams, enabling long-distance moisture transport from the North Atlantic and Red Sea, bringing moisture across arid regions [4].

The South Africa 2013 austral autumn event in May 2013 contributed to South Africa’s winter rainfall, with intense northward moisture flow originating from the South Atlantic and moving toward South Africa. The interaction between a subtropical high-pressure system and a low-pressure system over the continent intensified the event, highlighting ARs’ role in South African winter precipitation [7].

In April 2017, the MENA 2017 AR event impacting the region caused flooding and influenced snowmelt, especially in Iran. Moisture sources included the Red and Mediterranean Seas, with effects on areas such as Lake Urmia. The event also carried Saharan dust, affecting precipitation and ecosystems across long distances [29].

Illustrating rapid shifts from drought to flooding, the Mauritania 2019 event affected North Africa and the Middle East in March 2019. It resulted in severe flooding as far as Iran, causing extensive infrastructure damage and loss of life. This event exemplifies how climate extremes can intensify under changing climate conditions [5,52].

3. Results and Discussion

This chapter presents the results of the (1) statistical analysis of AR events over Africa using IPART and ARtracks. The accuracy of ERA5 data with regard to high-resolution RO measurements is evaluated with the comparison of RO and ERA5 data evaluated in (2).

3.1. Statistical Analysis of AR Events over Africa

This section presents results from the IPART and ARtracks analysis, highlighting AR frequency, monthly distribution, and hemispheric differences across the continent. We identified 1730 ARs impacting Africa between 2009 and 2019, with annual fluctuations shown in Figure 2. The number of ARs varies yearly, with the highest count in 2011 (174 ARs) and the lowest in 2009 (139 ARs). Overall, the number of ARs stays relatively constant over the period, shown by the annual average of 158 ARs.

Figure 2 illustrates both seasonal and interannual AR variability, showing event distribution by season each year. Here, and below, we denote the seasons as DJF (December, January, February), MAM (March, April, May), JJA (June, July, August), and SON (September, October, November). The seasonal averages for AR activity were 29 (JJA), 39 (SON), 47 (DJF), and 44 (MAM).

Notably, the years 2011, 2016, 2018 and 2019 show elevated AR counts. The DJF season is the most active season, with a peak in 2019 (59) and low points in 2014 (41) and 2015 (41), while MAM reaches its maximum in 2014 (56) and minimum in 2015 (37) and 2019 (37). JJA consistently records the least AR activity, ranging from 18 in 2013 to 40 in 2018, and SON exhibits moderate variability, with a peak in 2010 (51) and a low point in 2009 (29) and 2014 (29).

Average monthly AR distributions from 2009 to 2019 reveal distinct patterns, with peak activity in January (188 ARs), February (181 ARs) and March (198 ARs), a second peak in October (163 ARs), and a minimum in July (85 ARs), as shown in Figure 3.

AR activity gradually declines during April (156 ARs), May (154 ARs) and June (135 ARs). The heatmap in Figure 4 further highlights monthly AR frequencies, with darker shades indicating peak months.

DJF is the most active season, with peaks such as January 2011 (26 ARs), February 2018 (22 ARs) or January 2019 (23 ARs) shown in Figure 2. MAM also shows elevated activity, with the highest monthly count of 26 ARs occurring in March 2014. In contrast, JJA consistently exhibits the lowest AR counts, with a minimum of four ARs in July.

Southern and Northern Africa

In the statistical analysis, AR activity was separated by hemisphere at the equator, with results shown in Figure 5 for landfalling ARs from January 2009 to December 2019. The chart displays monthly AR activity in Northern Africa (blue line) and Southern Africa (green line).

Only ARs making landfall in Africa are included, excluding the Arabian Peninsula. Southern Africa shows consistently higher AR activity over the entire study period. There is a visible peak in January (142 ARs), high activity in February (119 ARs) and March (116 ARs), and a secondary peak in October (107ARs). In Northern Africa, high AR counts are observed from February (62 ARs) to May (66 ARs) with a secondary peak in October (56 ARs) and minimal activity in July and August, aligning with the study by Francis et al. (2022) [53], which highlights AR-driven moisture transport toward Europe.

In both regions, minimal activity is shown in July (Southern Africa: 69 ARs, Northern Africa: 16 ARs). Additionally, there is a secondary peak visible in October (Southern Africa: 107 ARs, Northern Africa: 56 ARs). The following sections provide a more detailed analysis of each region, with additional charts for further insight.

While Figure 5 shows the seasonality of AR landfalls in Southern Africa, where AR activity remains consistently higher than in Northern Africa, Figure 6 and Figure 7 provide a more detailed visualization of annual differences. They present heatmaps of monthly AR activity over Southern and Northern Africa, respectively.

AR in activity in Southern Africa is relatively high from October to March (see Figure 5 and Figure 6), but strong activity is observed from January to March, with highest peaks in January (e.g., 22 ARs in 2011 or 21 ARs in 2019) and February 2017 (23). In contrast, JJA (austral winter) shows minimal AR activity, with July typically recording the fewest events (e.g., only two in July 2014). This seasonal pattern underscores the role of ARs in Southern Africa’s wet season, contributing to summer precipitation, while winter remains drier with reduced AR influence [7]. Year-to-year variation is, again, evident, with higher AR activity in years like 2014 and 2019 and lower counts in 2009 and 2013.

Although the literature (e.g., [54]) often reports peak AR activity during austral winter (here understood as May-September), our 2009–2019 data show prominent activity during austral summer (DJF) and autumn (MAM), aligning with the region’s summer rainy season but diverging from some previous findings [6,7,55]. Our differing findings may stem from the limited existing research on Southern African ARs. Although our study period is comparatively short, we found noteworthy results, including the average number of 158 ARs per year and a small variation in the absolute number of landfalling ARs per year.

Northern Africa, situated between the subtropics and mid-latitudes, experiences most AR activity during DJF (boreal winter) and MAM (boreal spring), with lower overall frequency compared to Southern Africa. AR patterns are driven by mid-latitude, low-pressure systems that transport moisture from the Atlantic, which are crucial for this arid region’s water supply [5].

Figure 7 shows monthly AR characteristics in Northern Africa, with peaks in January 2019 (20 events), April 2017 (23 events), and March 2014 and 2019 (20 events) and the lowest AR activity in July and August.

While our findings agree with previous research [19,52,54,56] by identifying boreal winter as a peak period for AR activity, they differ by showing an additional peak in boreal spring (MAM). Other studies [19,52,54,56] observed the most AR activity during boreal autumn and winter. Our data, however, reveal a moderate peak in AR activity during October.

3.2. Comparison of RO and ERA5 Data

Here, we examine the relationship between IWV values from RO and ERA5 reanalysis data through case studies of six representative AR events in regions like Southern Africa, Middle East and North Africa (MENA), and West Africa. Through analysis of regression lines, Mean Biases, and RMSE, we assess the difference between ERA5’s and high-resolution RO data. It is important to note that while RO water vapor profiles possess high accuracy and vertical resolution, they tend to underestimate IWV when the profiles do not reach all the way down to the surface (as discussed in Section 1.3).

For the following comparisons, it is important to emphasize that both datasets are not “pure observations” and are subject to their own sources of uncertainty. The RO water vapor profiles are derived using a 1DVAR technique that relies on a background field (in this case, ERA5), while ERA5 itself is a model-based reanalysis that assimilates various observational data but also includes model assumptions and parameterizations.

When discrepancies arise between ERA5 and RO-derived IWV values, particularly in cases where ERA5 forecasts higher IWV than the collocated RO profiles, it is essential to consider the potential sources of these differences. The discrepancy could be due to either a moist bias in ERA5 or a dry bias in the RO retrievals. Without additional independent observations, such as collocated radiosonde measurements or other reliable observational data, it is not possible to definitively determine the source of the bias.

For example, in regions with high moisture content, ERA5 may overestimate IWV due to model parameterizations that struggle to accurately represent localized moisture transport or convection. In previous studies [23], we found via cross-comparisons that ERA5 data indeed show a tendency toward excess moisture compared to collocated ECMWF analysis data.

However, to provide a more robust interpretation of these discrepancies, future studies could benefit from the inclusion of collocated radiosonde data or other independent observations. These additional datasets would help to better quantify the biases in both ERA5 and RO-derived IWV values and provide a clearer understanding of the atmospheric conditions being studied.

Scatter plots and geospatial maps are utilized to compare IWV values from the two datasets at corresponding locations and times. The scatter plots illustrate the alignment between datasets, highlighting patterns and discrepancies. The red dashed line represents perfect alignment between ERA5 and RO datasets, while the green regression line denotes the line of best fit. The dots in the scatter plots represent RO data collected over the study period specified for each event in Table 1. Geospatial maps visualize the spatial distribution of moisture, enabling the identification of regions with consistent agreement and areas with notable deviations.

To streamline this section, we focus on two representative events—the South Africa 2009 event and the Morocco 2010 event. Detailed results for the other four events are summarized here, with full case studies provided in Appendix A for reference.

3.2.1. South Africa 2009 Event

This event delivered moisture from the South Atlantic and a remote South American source to South Africa, causing extreme rainfall at the Western Cape [8]. The scatter plot (Figure 8a) shows a high number of RO events (131) and a wide IWV range of ~3 to 43 kg m⁻², reflecting strong moisture transport surrounded by a dry environment. A spread in the data is evident at higher IWV values. At lower IWV values, below ~10 kg m⁻², good agreement between the RO data and the interpolated ERA5 IWV values is visible, as indicated by the close alignment of the data points (blue) along the regression line (green). This comparison yields an RMSE of 4.37 kg m⁻². The Mean Bias is −2.01 kg m⁻² and, together with the regression line slope of 0.82, this event indicates a tendency for ERA5 to show higher IWV values for RO. Supported by the findings in [23], we suppose that this discrepancy—which prevails throughout all of the events studied—is due to both a general underrepresentation of the true IWV in RO due to some missing data in the lowest levels, but is also due to an overrepresentation of humidity in ERA5.

The spatial map (Figure 8b) visualizes ERA5 IWV distribution overlaid with RO measurements (filled circles) and demonstrates the capability of the reanalysis dataset to capture large-scale moisture transport patterns associated with the AR event. The scale on the color bar indicates IWV values with yellow and green indicating lower and blue indicating higher values. The IWV values derived from the RO dataset are represented by the black-edged filled circles. Discrepancies in color indicate differences between the two datasets. On the edges of AR events higher discrepancy is expected due to sharp humidity gradients.

3.2.2. Morocco 2010 Event

The Morocco 2010 event was marked by extreme rainfall leading to widespread flooding and infrastructure damage [31]. The IWV in the scatter plot (Figure 9a) spans ~7 to 52 kg m⁻², underscoring the strong moisture transport associated with this AR, illustrated in Figure 9b. The Mean Bias of −2.33 kg m⁻² (the largest among all events) and RMSE of 4.41 kg m⁻² reflect moderate discrepancies. For this event, the ERA5 values are 122% of the RO values, which is shown by the slope of 0.82 that is the same as in the South Africa 2009 event. However, the scatter of the sample size of RO observations (44) is smaller than in the 2009 event.

Figure 9b illustrates high IWV over the Atlantic west of Morocco, which is captured well by ERA5, which shows extensive moisture transport toward the Moroccan coast. However, ERA5 reports systematically higher IWV levels in oceanic and northern Morocco regions compared to RO, which tends to retrieve drier profiles. IWV decreases inland, aligning well between datasets, though RO reports slightly lower values in some areas, particularly north of 35°N and south of 20°N.

The remaining four events—MENA 2010, South Africa 2013, MENA 2017, and Mauritania 2019—show similar trends, with ERA5 generally showing higher IWV compared to RO, while overall there is good agreement between ERA5 and RO. The slopes between 0.72 and 0.88 and consistent negative biases show systematic differences between the two datasets. RMSE values, ranging from 3.11 kg m⁻² (MENA 2010) to 4.53 kg m⁻² (Mauritania 2019), are moderate, indicating that ERA5 generally performs well when compared to RO but could benefit from further refinement.

Among the events, the South Africa 2009 event stands out with the largest number of RO events that provide a robust evaluation, while the Mauritania 2019 event is impacted by a decrease in RO observations during that time. The primary reason is the limited availability of RO satellite observations on that specific date. This event exhibits the weakest performance by ERA5 with the highest RMSE, the lowest slope, and the largest intercept. The MENA 2017 event demonstrates a strong linear relationship (highest slope) and minimal baseline offset (lowest intercept), suggesting ERA5 captures IWV variations well. The MENA 2010 event stands out with the strongest agreement between RO and ERA5 IWV, with the lowest RMSE and Mean Bias.

Overall, the analysis reveals that ERA5 and RO generally perform well in capturing IWV during moderate AR events (e.g., MENA 2010 and MENA 2017), where IWV ranges are narrower and biases are smaller. However, for stronger ARs with higher IWV values (e.g., Morocco 2010 and South Africa 2013), there are higher RMSEs and larger systematic biases, reflecting the combined uncertainties of both datasets rather than the errors of one specific dataset. Further details of these analyses are provided in Appendix A and Appendix B.

4. Conclusions

The findings of this research highlight the characteristics, seasonal trends, and regional differences in AR activity across Africa from 2009 to 2019. Additionally, this study evaluates the effectiveness of ERA5 compared to RO datasets in representing IWV during landfalling AR events. The key findings are summarized below:

• Annual Frequency and Distribution:
  A total of 1730 AR events made landfall in Africa during the study period, with a yearly average of 159 ARs. The years 2011 and 2018 showed the highest AR counts, with 174 and 171 events, respectively. The AR count was lowest for the whole study period in 2009, with 139 events.

2.

Seasonal Distribution and Monthly Trend:

Peaks of average monthly ARs counts for the whole continent occurred in January (188 ARs), February (181 ARs), March (189 ARs) and October (163 ARs). The most active season, with 47 ARs on average, was DJF, peaking in 2019 (59 ARs). Consistently, the least activity, with the lowest count in 2013 (18 ARs), was observed in JJA. SON showed moderate activity from 29 ARs in 2009 and 2014 up to 52 ARs in 2010. The second most active season, MAM, showed peak activity in 2014 (56 ARs) and a low in 2015 and 2019 (37 ARs).

3.

Regional Differences: Southern vs. Northern Africa:

Southern Africa experienced consistently higher AR activity throughout the year, peaking in austral summer (DJF). Northern Africa, however, saw a distinct seasonality, with AR events peaking in boreal winter (DJF) and spring (MAM).

4.

Event-Specific Insights:

The MENA 2010 event showed the strongest agreement between ERA5 and RO IWV values, with the lowest RMSE (3.11 kg m⁻²). Meanwhile, the Mauritania 2019 event demonstrated the weakest ERA5 performance, with the highest RMSE (4.53 kg m⁻²) and the largest intercept, indicating challenges in capturing extreme moisture conditions.

The observed discrepancies between ERA5 and RO IWV measurements highlight the importance of considering the uncertainties in both datasets when interpreting the results. While ERA5 may exhibit a potential moist bias in high-humidity regions, RO data also have limitations, particularly in the lower troposphere. The larger RMSE in certain cases should not be attributed solely to errors in ERA5 or RO but rather to the combined uncertainties of both datasets. Future studies could benefit from the inclusion of additional independent observations, such as radiosondes, to better quantify the biases in ERA5 and RO-derived IWV values.

5.

IWV and Pattern Consistency:

The analyzed AR events demonstrate a good overall agreement between ERA5 and RO IWV data. Acknowledging the fact that RO misses a part of the water vapor in the lowermost part of the profiles due to the RO signal often not penetrating all the way to the surface, and that ERA5 reanalyses tend to be too wet [23], we conclude that this systematic difference is due to both ERA5 and RO. Despite this, ERA5 effectively captured large-scale IWV patterns and high-moisture zones associated with AR events.

Comparisons between ERA5 and RO are currently somewhat limited due to a comparatively small number of RO profiles. However, expected increases in RO numbers in the future will allow for more detailed comparisons and for studies of AR events in other parts of the world.