Comparison of Moisture Sources and Sinks Estimated with Different Versions of FLEXPART and FLEXPART-WRF Models Forced with ECMWF Reanalysis Data

José C. Fernández-Alvarez aCentro de Investigación Mariña, Universidade de Vigo, Environmental Physics Laboratory (EPhysLab), Ourense, Spain
bDepartamento de Meteorología, Instituto Superior de Tecnologías y Ciencias Aplicadas, Universidad de La Habana, Havana, Cuba

Search for other papers by José C. Fernández-Alvarez in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0003-3409-6138
,
Marta Vázquez aCentro de Investigación Mariña, Universidade de Vigo, Environmental Physics Laboratory (EPhysLab), Ourense, Spain

Search for other papers by Marta Vázquez in
Current site
Google Scholar
PubMed
Close
,
Albenis Pérez-Alarcón aCentro de Investigación Mariña, Universidade de Vigo, Environmental Physics Laboratory (EPhysLab), Ourense, Spain
bDepartamento de Meteorología, Instituto Superior de Tecnologías y Ciencias Aplicadas, Universidad de La Habana, Havana, Cuba

Search for other papers by Albenis Pérez-Alarcón in
Current site
Google Scholar
PubMed
Close
,
Raquel Nieto aCentro de Investigación Mariña, Universidade de Vigo, Environmental Physics Laboratory (EPhysLab), Ourense, Spain

Search for other papers by Raquel Nieto in
Current site
Google Scholar
PubMed
Close
, and
Luis Gimeno aCentro de Investigación Mariña, Universidade de Vigo, Environmental Physics Laboratory (EPhysLab), Ourense, Spain

Search for other papers by Luis Gimeno in
Current site
Google Scholar
PubMed
Close
Free access

Abstract

Moisture transport and changes in the source–sink relationship play a vital role in the atmospheric branch of the hydrological cycle. Lagrangian approaches have emerged as the dominant tool to account for estimations of moisture sources and sinks; those that use the FLEXPART model fed by ERA-Interim reanalysis are most commonly used. With the release of the higher spatial resolution ERA5, it is crucial to compare the representation of moisture sources and sinks using the FLEXPART Lagrangian model with different resolutions in the input data, as well as its version for WRF-ARW input data, the FLEXPART-WRF. In this study, we compare this model for 2014 and moisture sources for the Iberian Peninsula and moisture sinks of North Atlantic and Mediterranean. For comparison criteria, we considered FLEXPARTv9.0 outputs forced by ERA-Interim reanalysis as “control” values. It is concluded that FLEXPARTv10.3 forced with ERA5 data at various horizontal resolutions (0.5° and 1°) represents moisture source and sink zones as represented forced by ERA-Interim (1°). In addition, the version fed with the dynamic downscaling WRF-ARW outputs (∼20 km), previously forced with ERA5, also represents these patterns accurately, allowing this tool to be used in future investigations at higher resolutions and for regional domains.

Significance Statement

The FLEXPART dispersion model forced with ERA5 reanalysis data at various resolutions represents moisture source and sink zones compared to when it is forced by ERA-Interim. When the Weather Research and Forecasting Model is used to dynamically downscale ERA5, FLEXPART-WRF can also represent moisture sources and sinks, allowing this tool to be used in future investigations requiring higher resolution and regional domains and on regions with a predominance of complex orography due to its ability to represent local moisture transport.

© 2023 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: José C. Fernández-Alvarez, jose.carlos.fernandez.alvarez@uvigo.es

Abstract

Moisture transport and changes in the source–sink relationship play a vital role in the atmospheric branch of the hydrological cycle. Lagrangian approaches have emerged as the dominant tool to account for estimations of moisture sources and sinks; those that use the FLEXPART model fed by ERA-Interim reanalysis are most commonly used. With the release of the higher spatial resolution ERA5, it is crucial to compare the representation of moisture sources and sinks using the FLEXPART Lagrangian model with different resolutions in the input data, as well as its version for WRF-ARW input data, the FLEXPART-WRF. In this study, we compare this model for 2014 and moisture sources for the Iberian Peninsula and moisture sinks of North Atlantic and Mediterranean. For comparison criteria, we considered FLEXPARTv9.0 outputs forced by ERA-Interim reanalysis as “control” values. It is concluded that FLEXPARTv10.3 forced with ERA5 data at various horizontal resolutions (0.5° and 1°) represents moisture source and sink zones as represented forced by ERA-Interim (1°). In addition, the version fed with the dynamic downscaling WRF-ARW outputs (∼20 km), previously forced with ERA5, also represents these patterns accurately, allowing this tool to be used in future investigations at higher resolutions and for regional domains.

Significance Statement

The FLEXPART dispersion model forced with ERA5 reanalysis data at various resolutions represents moisture source and sink zones compared to when it is forced by ERA-Interim. When the Weather Research and Forecasting Model is used to dynamically downscale ERA5, FLEXPART-WRF can also represent moisture sources and sinks, allowing this tool to be used in future investigations requiring higher resolution and regional domains and on regions with a predominance of complex orography due to its ability to represent local moisture transport.

© 2023 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: José C. Fernández-Alvarez, jose.carlos.fernandez.alvarez@uvigo.es

1. Introduction

Currently, for the fields of hydrology, climatology, and meteorology, it is necessary to understand the origin of humidity and precipitation that occurs over a given region, especially on continents, in which water resources play a vital role (Randhir 2012). Considering that approximately 90% of the water in the atmosphere comes from evaporation over the oceans, lakes, and other open water bodies, its atmospheric transport plays an important role in the precipitation component of the hydrological cycle, allowing redistribution of water toward the land (Quante and Mathias 2006). Therefore, to understand moisture transport processes, it is necessary to know how water vapor is distributed in the atmosphere, considering its concentration is highly variable in space and time (Gimeno et al. 2010a). In this sense, the moisture transport and changes in the source–sink relationship can play a critical role in the hydrological cycle, allowing for the analysis of variations in the relative importance of oceanic sources versus terrestrial sources in continental precipitation at a large scale (Gimeno et al. 2012). On the other hand, in the context of climate change, it is important to study future changes associated in the source–sink relationship and how it may influence continental precipitation.

Different methods have been developed to identify atmospheric moisture sources and their related sinks; they can be classified into numerical water vapor tracers, analytical models, and physical water vapor tracers using isotopes. Gimeno et al. (2012, 2020) demonstrated their validity, specific uses, advantages, and disadvantages. Various studies on moisture source–sink assessments have been conducted at global scales. In the last decades, several authors have investigated moisture transport in different regions. A complete summary of the studies is shown in Gimeno et al. (2020), including investigations related to extreme events such as droughts or floods and meteorological or circulation systems. Additionally, numerical methods have been increasingly used. Such methods can be classified into Eulerian and Lagrangian; the former deals with the water balance at fixed locations as time varies, while the latter is based on studying the water vapor budget of air parcels as they travel either forward or backward in time and space (Stohl and James 2005). The Lagrangian Flexible Particle (FLEXPART) dispersion model (Stohl and James 2004), in its several versions, has been widely used in many studies to determine moisture sources and sinks usually fed with the three-dimensional (3D) ERA-Interim global atmospheric reanalysis data (ERA-I; Dee et al. 2011) from the European Centre for Medium-Range Weather Forecasts (ECMWF), covering the period from January 1979 to August 2019 with a 1° horizontal spatial resolution. However, this reanalysis has been superseded by the ERA5 reanalysis (Hersbach et al. 2020).

A current challenge for the coming years in the atmospheric moisture transport field requires advances in the evaluation of the moisture sources, which has been limited mainly due to the grid scale of the model results (Gimeno et al. 2020). These advances allow for a better understanding of the role of moisture transport in precipitation distribution. As such, the new ERA5 reanalysis data have a higher spatial (0.25°, 0.5°, and 1°) and temporal resolution (1 h), including an improved representation of the troposphere, a better global balance of precipitation and evaporation, and more consistent coverage of sea surface temperature and sea ice (Hersbach et al. 2020). In addition, this dataset covers a longer period, from 1950 to the present, with daily updates being available 5 days behind real time. These improvements may better represent the source–sink relationships involved in the atmospheric branch of the hydrological cycle; however, it is necessary to compare this fact with respect to the previous reanalysis products such as ERA-I. Furthermore, at the time of submission of this work, no studies have compared moisture sources and sinks using ERA5 data to feed any version of the FLEXPART model to our best knowledge.

If the further representation of the physical processes involved in moisture transport at a regional scale is required, it is necessary to increase the resolution of the computational domain. These more detailed results can be achieved feeding the Lagrangian model with a dynamic mesoscale model for a regional domain. Recently, the use of the FLEXPART model adapted for input data from the Weather Research and Forecasting (WRF) Model (Brioude et al. 2013) but forced with National Centers for Environmental Prediction (NCEP) reanalysis data (Kalnay et al. 1996) was used to analyze the moisture sources associated with a rainfall event using backward trajectories over Japan in July 2020 (Zhao et al. 2021). Furthermore, Cloux et al. (2021) found that this model performed well in identifying local and mid-distance sources involved in extreme precipitation events in the Mediterranean area but performing simulations with WRF outputs forced by ERA-I that is no longer updated. Both studies are highly local and focused on the effects of the precipitation associated with meteorological systems. However, it is still necessary to evaluate the ability of this model to represent large-scale sources and sinks of moisture.

Therefore, the objective of this investigation was to compare the representation of moisture sources and sinks for ERA5 reanalysis data with different configurations of the Lagrangian models FLEXPART and FLEXPART-WRF for 2014, carried out for the North Atlantic region and most of Europe, taking into account that the version of WRF used is optimized for the Iberian Peninsula (IP) (Insua-Costa and Miguez-Macho 2018; Insua-Costa et al. 2019), and the analyzed region includes two of the main global oceanic moisture sources, the Mediterranean Sea (MED) and the North Atlantic Ocean (NATL) (Gimeno et al. 2010a).

2. Materials and methods

a. Configurations and data used to force the different FLEXPART model versions

The outputs of the FLEXPARTv9.0 Lagrangian model (Stohl 1998; Stohl and Thomson 1999; Stohl et al. 2005) forced by ERA-I reanalysis data (Dee et al. 2011) from the ECMWF with a 1° horizontal resolution and 61 vertical levels (herein, FLEX-ERA-I.1) are considered as our “control” values. The simulation was carried out for a global domain, tracking every 6 h forward in time 2.0 million air parcels (or particles) into which the atmosphere is evenly divided. This number of particles is chosen to have at least one particle at each point of the grid and most vertical levels. This configuration is the same as the one widely used in previous research to represent the behavior of moisture sources and sinks at global or regional scales (Gimeno et al. 2020, and references therein). This FLEXPART version (v9.0) allows working with input data from the ECMWF Integrated Forecast System, such as ERA-I, but does not allow the newest ERA5 (Hersbach et al. 2020). Pisso et al. (2019) described the physical characteristics, parameterizations, and improvements of the different FLEXPART models over time. FLEXPARTv9.0 includes a global land-use inventory, allowing accurate dry deposition calculations everywhere on the globe, distinguishing between in-cloud and below-cloud scavenging for washout, relying on simple diagnostics for clouds based on gridscale relative humidity. The physics of the FLEXPARTv9.0 model is described by the zero-acceleration scheme and a set of parameterizations. For instance, Hanna’s (1982) parameterization scheme is used for wind fluctuations, the frictional velocity is computed using surface stresses, and sensible heat fluxes is considered in the boundary layer parameterization. Meanwhile, the profile method (Berkowickz and Prahm 1982) is applied in the absence of frictional velocity information. Turbulent motion is parameterized considering a Markov process based on the Langevin equation (Thomson 1987). Emanuel and Živković-Rothman’s (1999) one-dimensional convection model is used as a convection scheme. Radioactive decay and dry and wet deposition are also considered in the physics of the model.

The latest version of this Lagrangian model, FLEXPARTv10.3, was updated to run with the ERA5 dataset (among other changes), which is particularly important given that ERA-I has no longer been updated since August 2019. The new version of this Lagrangian model (Pisso et al. 2019) presents modifications with respect to the previous v9.0. A new scheme applying more realistic skewed rather than Gaussian turbulence statistics in the convective atmospheric boundary layer has been incorporated. Furthermore, the wet deposition scheme for aerosols was revised (Grythe et al. 2017), introducing dependencies on aerosol size and precipitation type (rain or snow) and distinguishing between in-cloud and below-cloud scavenging. Moreover, the model reads 3D cloud water fields from meteorological input files.

Considering that this work compares the ability of the new FLEXPARTv10.3 to reproduce moisture sources and sinks forced by several horizontal resolutions of the new ERA5, it is necessary to highlight some improvements of ERA5 with respect to ERA-I, which are considerable in the troposphere (Hersbach et al. 2020); these include changes in the physical parameterizations of subgrid-scale processes and the formulation of the organized deep entrainment in the convection parameterization scheme (Bechtold et al. 2008); the moisture-convergence-based entrainment formulation was replaced with environmental relative-humidity-dependent entrainment. A complete evaluation of the ERA5 reanalysis is presented by Hersbach et al. (2020). According to Hersbach et al. (2020), the improvements in precipitation field occur both over the extratropical regions, tropical oceanic zones [where correlations were already quite high in ERA-I compared to the Tropical Rainfall Measuring Mission (TRMM/3B43)], and over the tropical landmasses, where ERA-I showed particularly low correlations). Moreover, global correlations with data from the Global Precipitation Climatology Centre (GPCP) at a 2.5° resolution (from 1979 to 2018) are also improved with ERA5 (77% versus 67% for ERA-I). In addition, ERA5 provides data with a higher vertical resolution, with 137 vertical hybrid sigma/pressure levels from 1013.25 to 0.01 hPa; ERA-I only has 61. Moreover, ERA5 provides a larger set of different horizontal spatial resolutions (Hersbach and Dee 2016), including 0.25°, 0.5°, and 1°. The main changes in ERA5 compared to ERA-I that can improve the wet and dry biases (Hersbach et al. 2020) are shown in Table 1.

Table 1

Main changes in ERA5 vs ERA-I improving the wet and dry biases (from Hersbach et al. 2020).

Table 1

Here, the FLEXPARTv10.3 model has been fed every 6 h with two datasets of ERA5 reanalysis data at different resolutions: first, for a direct comparison with our control experiment, with ERA5 at 1° (hereafter named as FLEX-ERA5.1), and then, to increase the resolution of the Lagrangian outputs, with ERA5 at 0.5° (herein denoted as FLEX-ERA5.05). It should be noted that vertical resolution improved using ERA5 compared to ERA-I, and consequently the FLEX-ERA5.1 and FLEX-ERA5.05 versus the control experiment FLEX-ERA-I.1.

The FLEXPART model needs to maintain the atmospheric mass (one particle per vertical level is required) during the simulation; this must be considered in terms of the number of particles to be simulated. As ERA5 has a higher vertical resolution than ERA-I (more than double, 137 versus 61 levels), therefore the parcels in which the atmosphere is divided must be greater than in the FLEX-ERA-I.1 control simulation. Therefore, for the FLEX-ERA5.1 experiment, the atmosphere is homogeneously divided and evenly distributed into approximately 9 million particles, and 30 million for FLEX-ERA5.05. It is important to note that the data download time becomes slower as the horizontal and vertical resolution increases and that the run time also increases as a function of the number of particles into which the atmosphere is divided, which is why the global experiment with ERA5 at 0.25° is not operational. The best ratio between time, computational power, and capacity for storing the model outputs is decisive when deciding on the experiments to be performed.

Although the experiments explained above have been run on a global scale, it is possible to obtain resolution gains and improve the physical characterizations of the results by using input data on a regional scale. With this aim, a dynamic downscaling was carried out using the WRF-ARWv3.8.1 model (Skamarock et al. 2008) forced every 6 h with ERA5 reanalysis data at 0.25° in latitude and longitude. Figure 1a shows the regional domain for the study, 115°W–42°E and 19°S–59°N. WRF outputs were downscaled to 0.18° of the horizontal spatial resolution and saved every 6 h. A spinup was carried out for a month before the beginning of the study period. These simulations were carried out in two intervals using the restart mode of the WRF-ARW. We used the WSM6 microphysics scheme (Hong and Lim 2006), the Yonsei University PBL scheme (Hong et al. 2006), the revised MM5 surface layer scheme (Jimenez et al. 2012), the United Noah Land Surface Model (Tewari et al. 2004), the RRTMG shortwave and longwave schemes (Iacono et al. 2008), and the Kain–Fritsch ensemble cumulus scheme (Kain 2004). Spectral nudging of waves longer than 1000 km was activated to avoid distortion of the large-scale circulation within the regional model domain due to the interaction between the model’s solution and the lateral boundary conditions (Miguez-Macho et al. 2004). The selection criteria for these parameterizations are based on the fact that they have been evaluated in several investigations for the region (e.g., Insua-Costa and Miguez-Macho 2018; Insua-Costa et al. 2019).

Fig. 1.
Fig. 1.

(a) Domains for WRF-ERA5 (red) and FLEX-WRF (green) simulations. The individual target regions under the study are colored: the Mediterranean Sea (MED) is shown in blue, the Iberian Peninsula (IP) in pink, and the North Atlantic Ocean (NATL) in dark purple. (b) Orography of the region, where the WRF simulations are performed, was taken from the HydroSHEDS project (Lehner et al. 2008).

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

The regional WRF-ARW outputs using ERA5 (herein WRF-ERA5) were used to run the appropriated version of FLEXPART adapted for WRF in its v3.3.2 version (Brioude et al. 2013). WRF-ERA5 outputs have 40 vertical levels in sigma coordinates from the surface to approximately 50 hPa (pressure top to use in the model) and 480 × 780 grid nodes (∼0.18°). The FLEXPART-WRF used these 40 levels reducing the computational cost maintaining the atmosphere properties since the primary changes in humidity occur down the tropopause. The simulated area has 400 × 777 grid points (smaller in size than WRF outputs), where the particles are released. In addition, for FLEXPART-WRF, we used Hanna’s scheme for turbulence parameterization (Hanna 1982), with a convection scheme activated. In this simulation, to maintain the distribution of atmospheric mass, the atmosphere was homogeneously divided into 2 million air parcels (or particles), which were subsequently moved forward in time by the model along the whole study period. The outputs from this experiment are denoted as FLEX-WRF in the text. Table 2 summarizes the FLEXPART versions of the model, input and output data characteristics, the particles moved by the model, vertical atmospheric levels, pressure top, and number of levels up to 50 hPa (NL50 hPa).

Table 2

Summary of the models and input database used for the different experiments. Characteristics of the input and output model data, resolution, number (in millions) of released particles, vertical atmospheric levels, the top pressure level, and number of levels up to 50 hPa are listed for each experiment.

Table 2

b. Moisture sources and sinks used for the comparison

The selection criteria for the study regions were based on two reasons: first, that the North Atlantic and European areas encompass two of the main global oceanic moisture sources, the NATL and the MED (as defined in Gimeno et al. 2012), and second, due to the availability of a correctly regionalized WRF-ARW Model for the IP (Insua-Costa and Miguez-Macho 2018; Insua-Costa et al. 2019), an area located between these both sources and a sink of moisture from both in different seasonal periods (Gimeno et al. 2010a,b). Focusing on these three areas (included in Fig. 1a), we have compared the ability of the new FLEXPARTv10.3 fed with ERA5 and the FLEXPART-WRFv3.3.2 configurations to represent the atmospheric moisture sources and sinks.

The NATL source is considered a dominant oceanic source that contributes moisture to continental precipitation at both sides of the basin, from North to South America, and over Europe and northern Africa (Gimeno et al. 2010a). The importance of this source has been well documented in several previous analyses, e.g., for Central America (Durán-Quesada et al. 2010), South America (Drumond et al. 2008), or Europe (Gimeno et al. 2010b). Moreover, the NATL contributes to the North and South American monsoon systems, as well as to the Atlantic intertropical convergence zone (ITCZ) (Castillo et al. 2014; Drumond et al. 2011a). According to Gimeno et al. (2010a), the moisture provided from NATL to be transported in the atmosphere increases during winter and decreases strongly in summer, although it does not show many variations in its size and position throughout the year. Many authors have identified the NATL as an important moisture source for Europe, mainly in autumn and winter, as well as in summer months with less influence (e.g., Sodemann and Stohl 2013; Gómez-Hernández et al. 2013). In addition, it has also been found that moisture transport toward Europe from the North Atlantic is strongly influenced by the cyclonic activity and atmospheric rivers (Lavers and Villarini 2013).

The MED, positioned at the border between the tropical climate zone and the midlatitude climate belt, presents a large meridional gradient. Its particular geography, a completely closed basin connected to the Atlantic Ocean through the narrow Gibraltar Strait, high mountain ridges surrounding it (its Alps reach 4800 m), islands, peninsulas, and many regional seas and basins, determines a complicated land–sea distribution pattern with a large spatial variability of sea and atmospheric circulation with many subregional and mesoscale features (Lionello et al. 2006). MED plays an important role in the study area in terms of atmospheric moisture transport for precipitation. During boreal summer (when its contribution is higher), it provides moisture to its surroundings, reaching all directions, extending into northern Europe, northeast Africa, and the Middle East. During autumn and winter, its contribution as a moisture supplier for continental areas is reduced, similar to the IP (Gimeno et al. 2010a). The significance of the MED as a moisture source lies in the fact that, for some regions adjacent to it, it is the single major oceanic source of moisture for precipitation (Schicker et al. 2010; Nieto et al. 2010; Drumond et al. 2011b; Gómez-Hernández et al. 2013).

The IP is in southwestern Europe, surrounded by the Mediterranean Sea to the east and the Atlantic Ocean to the west. The precipitation regime in the north and west of the IP is strongly affected by the mean annual cycle of the Atlantic storm track and its variability, whereas in the interior and east of the IP, large-scale synoptic systems share importance with convective precipitation (Trigo et al. 1999, 2000). The main moisture sources that affect the IP are the tropical–subtropical region of North Atlantic Ocean and Mediterranean Sea (a more local source). In addition, the recycling process is a characteristic of the IP, predominantly in the east and center of the region and less relevant in the west and north (Gimeno et al. 2010a; Rios-Entenza and Miguez-Macho 2014; Rios-Entenza et al. 2014).

The period selected for the comparative analysis was 2014. We studied the averaged conditions throughout this year. Despite the short temporal time frame, this can be considered a standard year because the major modes of variability over the region did not show extremes in their phases, such as the North Atlantic Oscillation or El Niño–Southern Oscillation.

c. Identification of the source and sinks of moisture

The particles modeled from the different versions of FLEXPART can be followed backward or forward in time along their trajectories to identify moisture sources and sinks, respectively. For that purpose, the particles over a selected region were followed normally for 10 days, which is considered the average residence time of water vapor in the atmosphere (Numaguti 1999; van der Ent and Tuinenburg 2017).

For each particle, the moisture variation along each trajectory every 6 h was computed as
(ep)=m(dqdt)
where e is the evaporation from the environment, p the precipitation, m the mass of the parcel, and q the specific humidity. Once the individual (ep) computations for all the parcels were computed for all the trajectories, the total surface freshwater flux at each grid cell (EP) was computed by adding the contribution by all parcels residing over a gridded area (A) at a specific time. The total budget is computed as
(EP)=k=1N(ep)A,
where E represents total evaporation, P total precipitation, and N is the total number of particles over A.

It is necessary to consider that the moisture sources for a region are defined as those areas where evaporation dominated over precipitation (i.e., EP > 0) when the backward in time is performed. Otherwise, to find the moisture sinks of air masses leaving a region, it is necessary to detect where they show a net loss of moisture, that is, where precipitation dominates over evaporation, (EP) < 0. For a better representation and interpretation of (EP) field in the case of sinks (precipitation), it is multiplied by −1 to represent positive rainfall values on the maps, and we will refer to it along the manuscript hereafter as (PE) > 0.

Considering the methodology presented above, in the interval of the 10 days of the trajectories there are 40 values of (ep), four values every 6 h for each day and particle. For the calculation of the daily pattern (EP)[d] [d = 1–10 days] according to Eq. (2), the changes in specific humidity (q) are considered along these four values during one day for each particle [Eq. (1)], and the final daily (ep) value is assigned to the last geographical position of the particle. This ensures that all contributing particles in a given area are taken into account. Subsequently, the sum of the daily values is divided by the area of the considered grid. The pattern integrated for the 10 days, (EP)[1–10], is the sum of (EP)[d] from days 1 to 10. As (EP)[1–10] pattern is determined for the 365 days of the year 2014, the annual and seasonal (EP)[1–10] patterns could be calculated as means of convenient periods [the whole year, or January–March (JFM), April–June (AMJ), July–September (JAS), and October–December (OND)]. Both methodologies (back and forward performances) and (EP) field representation, for individual days and/or integrated along the whole trajectories, have been used in numerous studies (e.g., Castillo et al. 2014; Drumond et al. 2008; Nieto et al. 2006; Vázquez et al. 2020).

In this work we compare (i) the moisture sources for a target region, the IP, and (ii) the sinks of moisture for the NATL and MED basin sources. Therefore, backward and forward modes were used, respectively.

d. Comparison methodology

The annual and seasonal comparison (increases/decreases) between the dataset obtained in this work was carried out using the ERA-I dataset and derived fields as the control and for 2014. The boreal seasonal periods were selected as in Gimeno et al. (2010b): winter (JFM), spring (AMJ), summer (JAS), and autumn (OND).

First, we compare differences in the integrated water vapor transport [IVT, see Eq. (3)] fields among ERA5, WRF-ERA5, and the control ERA-I datasets. This was done to determine if the WRF Model correctly represents the variables involved in the moisture transport, as the specific humidity and the zonal and meridional wind. The IVT was calculated as
IVT=uq2+υq2,
uq=1gpspuqdp,
υq=1gpspυqdp,
where g is the gravitational acceleration, q is the specific humidity, ps is the surface pressure, p pressure on the top, and u and υ are the zonal and meridional wind, respectively (Lavers et al. 2012; Zhang et al. 2019). The pressure levels used for IVT calculation were from 1000 to 300 hPa (Ramos et al. 2018).

Later, we compare the differences in the moisture fields (sources and sinks) obtained from the different FLEXPART models used in this work. The comparison was made for the (EP) field integrated during the 10 days [denoted as (EP)[1–10]], and on individual days (d): 1, 3, 5, and 10 [denoted as (EP)[d]]. In addition, to analyze whether the representation of moisture sources and sinks are correct, the divergence of the IVT (DIVT) was also checked.

The statistics used (Brown et al. 2013) were the mean absolute error (MAE), root-mean-square error (RMSE), Pearson’s correlation (R), bias (B), and standard deviation (STD).

The MAE [Eq. (6)] measures how far two variables or fields are; the closer values are to zero, the more accurate the simulation will be:
MAE=i=1n|xiyi|n.
The RMSE [Eq. (7)] can quantify the magnitude of the deviation of the simulated values concerning the control; when the value of RMSE is equal 0 the simulation is considered perfect:
RMSE=i=1n(xiyi)2n.
In Eqs. (6) and (7) [and in Eqs. (8)–(10)], xi represents the fields from FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF and yi the control values from FLEX-ERA-I.1; n is the number of points.
The R is an index that measures the degree of covariation between different linearly related variables [Eq. (8), where x¯i and y¯i are the mean values of xi and yi]. The correlation between two variables X and Y is positive; when one of them increases, the other increases. The closer to one is the value of the correlation coefficient, the more similar the behavior between both variables:
R=i=1n(xix¯i)(yiy¯i)i=1n(xix¯i)2(yiy¯i)2.
The B [Eq. (9)] provides the difference between a field of values and the control one; the closer the B values are to zero, the more accurate the simulation. The B < 0 values indicate a decrease with respect to the control, while B > 0 indicates an increase:
B=i=1n(xiyi)n.
The STD [Eq. (10)] provides a variability measure in the same units as the quantity being characterized:
STD=11ni=1n(xix¯i)2.
Finally, the coefficient of determination (R2) is used to eliminate a possible exaggeration of the similarities between different sets of results shown by Pearson’s coefficient. In addition, the coefficient of variation (VAR = STD/mean) was used to analyze the relationship between the size of the mean and the variability of the variable. It is important to note that when the mean is very close to zero when calculating STD/mean, the value increases, losing the meaning of the coefficient of variation, and therefore it does not necessarily imply a large dispersion of data.

3. Results

The results of the comparison of each configuration with respect to FLEX-ERA-I.1 will be presented following the conceptual diagram in Fig. 2. Initially, the variables involved in moisture transport were analyzed [precipitation (P), evaporation (E), total column water (TCW), IVT, and specific humidity at 850 hPa (q850)]. Later the patterns for moisture sources and sinks (EP fields) were compared for each target region considered in the study. After that, an analysis of the representativeness of each configuration were carried out using Taylor diagrams. In addition, the MAE and RMSE were determined to know the approximate errors that are made in the use of each configuration in the representation of the patterns of sources and sinks with respect to the control values. Finally, and taking into account the results obtained in the previous steps, the conclusions that make up this analysis are drawn.

Fig. 2.
Fig. 2.

Conceptual diagram used in the research.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

a. Comparison of fields related to moisture transport

Figure 3 shows the IVT fields from the WRF-ERA5 (at 0.18°), ERA5 (at 0.25°), and ERA-I (control values, at 1°) datasets. The results for ERA5 and WRF-ERA5 match with ERA-I, showing similar areas of maxima and minima values; they accurately represent the known movement of the North Atlantic anticyclone to the west in summer or eastward in autumn, and the IVT maximum toward Europe during winter coming from the NATL source (Gimeno et al. 2010b). In addition, the ITCZ was also captured as an area of IVT maxima during all seasons that influence Central America, the Caribbean Sea, and South America.

Fig. 3.
Fig. 3.

Annual and seasonal IVT module (colors; kg m−1 s−1) and direction (arrows; kg m−1 s−1) for ERA5, WRF-ERA5, and ERA-I for 2014 at 0.25°, 0.18°, and 1° resolutions, respectively. The fields displayed correspond to (a)–(c) winter (JFM), (d)–(f) spring (AMJ), (g)–(i) summer (JAS), (j)–(l) autumn (OND), and (m)–(o) annual.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

The higher resolution of WRF-ERA5 and ERA5 captured greater details about the IVT field over areas with significant moisture transport, for instance over the east coast of South America in the Amazon and over the North Atlantic at 30°N. However, WRF-ERA5 increases the IVT (comparing with ERA-I) over certain areas with complex orography (e.g., the Andes, the Rocky Mountains, and southeast of the African continent) and at the edges of the domain. Similar to Rahimi et al. (2022), it seems that WRF simulates a higher IVT than ERA5 in both the ERA5- and global climate models–driven experiments. This could be related to the conditions provided by the WRF Model for the simulations that generate certain uncertainties (Warner et al. 1997); in this case, using local parameterizations and to the initial and boundary conditions, respectively. Visual analysis of the IVT field shows that both products match for the region and therefore allow their use as forcers for the FLEXPARTv10.3 and FLEX-WRF models.

On the other hand, a comparison of the fields related to the water cycle (P, E, IVT, TCW, and q850) from ERA5 and ERA-I reanalyses was carried out, and a Student’s t test was used to determine the areas where exits significant differences (at significance level of 5%). Changes and improvement of the ERA5 reanalysis over the areas that act as moisture sources or sinks for our target areas (all north to 10°N) are the key to achieve the better source–sink relationship. The ERA5 total precipitation (P) pattern shows a decrease in the tropical region and intertropical convergence zone and a relative increase for the midlatitude regions (especially in winter and spring) compared with ERA-I (Fig. S1 in the online supplemental material). These results are due to the fact that ERA-I overestimates (underestimates) deep convection and moisture flux convergence over the tropics (midlatitudes), leading to excessive (less) precipitation (Nogueira 2020). Both cases are significantly improved in ERA5, because of better parameterizations and its higher resolution (see Table 2). Significant differences (at 95%) appear over the Caribbean Sea and tropical region; however, in midlatitudes no significant changes are observed. The evaporation (E) field comparison shows a general increase for ERA5, but not significant, and a significant decrease over the western middle Atlantic Ocean basin in summer and autumn (Fig. S2). It is notable that the Mediterranean Sea and North Atlantic regions show an increase from 1 to 2 mm day−1. Nogueira (2020) showed that surface evaporation is slightly higher in ERA5 compared to ERA-I, showing a behavior inversely proportional to precipitation. The total column water (EP) corresponding to ERA5 with respect to ERA-I shows a general decrease throughout the domain, except for the mountainous regions of the Andes Mountains with statistically significant differences (Fig. S3). Notable decreases are observed in the Mediterranean Sea and the North Atlantic, mainly in the seasons of spring, summer and autumn (statistically significant at 95%). In general, the IVT calculated from ERA5 presents lower values than from ERAI; however, areas of increase are observed near the coasts of Africa and over the continent for the winter and annual periods (Fig. S4). In addition, the specific humidity field at 850 hPa presents lower values for ERA5 than for ERA-I, but with the greatest differences over the tropical regions of the Northern Hemisphere and over the South Atlantic (Fig. S5). Maximum differences stand out near the east coast of South America and west of Africa around 10°S. Nogueira (2020) showed that, overall, ERA5 improved the representation of moisture sink/source patterns, primarily over tropical oceans.

b. Representation of the moisture sources for the IP

In this section, the sources of moisture for the IP [(EP)[1–10] > 0 fields], from the FLEX-WRF, FLEX-ERA5.05, and FLEX-ERA5.1 models were compared with those from FLEX-ERA-I.1 (the control) to demonstrate that the different configurations forced with ERA5 represent the moisture sources and results using the ERA-I reanalysis found by previous research as in Gimeno et al. (2010a).

Figure 4 represents the moisture sources for the IP determined from the outputs of each FLEXPART model in the study (first to fourth columns). The fifth column represents the IVT (arrows) and its divergence (colored areas), computed using the ERA-I dataset. In general, there is suitable correspondence between the source zones and areas of IVT divergence. To check the robustness of the results, different statistical parameters were calculated to compare the different configurations with the control experiment. The STD, correlation coefficient, and B are plotted in Fig. 5 for each season, while Fig. 6 shows the MAE and RMSE.

Fig. 4.
Fig. 4.

Moisture sources patterns [(EP)[1–10] > 0; mm day−1] for the IP from the Lagrangian outputs of FLEX-ERA5.05, FLEX-ERA5.1, FLEX-WRF, and FLEX-ERA-I.1. The right column represents the IVT field (arrows; kg m−1 s−1) and its divergence (contours; mm day−1) from ERA-I at 1°. The fields displayed correspond to (a)–(e) JFM, (f)–(j) AMJ, (k)–(o) JAS, (p)–(t) OND, and (u)–(y) annual.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

Fig. 5.
Fig. 5.

Taylor diagrams to compare the (EP)[1–10] > 0 fields of FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF with respect to the control field FLEX-ERA-I for the moisture sources for the IP in terms of B, correlation coefficient, and STD. (a) JFM, (b) AMJ, (c) JAS, (d) OND, and (e) annual. The letters A, B, and C correspond to the configurations FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF, respectively.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

Fig. 6.
Fig. 6.

MAE and RMSE for the moisture sources patterns of FLEX-ERA5.05, FLEX-ERA5.1, FLEX-WRF, and FLEX-ERA-I for the (a),(b) IP, (c),(d) MED, and (e),(f) NATL. The red, blue, and green bars correspond to FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF, respectively.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

In winter (JFM), the main source of moisture for the IP, the NATL (Gimeno et al. 2010b), is well represented by all the configurations; in general, they all represented a similar spatial pattern over the North Atlantic Ocean and the IP surroundings; the recycling process over the IP was also captured. The small STD confirmed this (<0.2 mm day−1) and R values ranging 0.45–0.5 (Fig. 5a). The B showed a general decrease, with values of 0.07, 0.09, and 0.11 for the FLEX-ERA5.1, FLEX-WRF, and FLEX-ERA5.05 experiments, respectively (see Fig. 5a). FLEX-ERA5.1 showed the lowest MAE, with similarly higher values for the other two experiments FLEX-ERA5.05 and FLEX-WRF (Fig. 6a). The RMSE (Fig. 6b) showed values ∼0.2 mm day−1 for FLEX-WRF and FLEX-ERA5.1 but slightly higher for FLEX-ERA5.05. The RSME weighs the maximum errors and, therefore, indicates the largest errors for each configuration. The higher RMSE values observed in FLEX-ERA5.05 can be related to the smoother pattern over the Atlantic corridor (see Fig. 4).

In spring (AMJ), there is suitable agreement in the pattern of the three configurations under study (Figs. 4f–i), as shown by high R (around 0.7) and STD values (around 0.2–0.4 mm day−1) in Fig. 5b. The lowest STD was obtained for FLEX-ERA5.05 and the highest for FLEX-ERA5.1. Again, the B indicates a decrease of the sources compared to the control values for the three configurations. However, B differs between them, showing the lowest values for FLEX-ERA5.1 (∼0.02 mm day−1) followed by FLEX-WRF ∼0.06 mm day−1 and FLEX-ERA5.05 with ∼0.08 mm day−1. Regarding the MAE, FLEX-ERA5.1 showed a lower value than the remaining configurations, followed by FLEX-ERA5.05 and FLEX-WRF, which showed very little difference. RMSE differs in this order (see Figs. 6a,b): the highest value was found for FLEX-ERA5.05 (∼0.4 mm day−1) but with similar values for FLEX-ERA5.1 and FLEX-WRF close to 0.38 mm day−1.

In summer (JAS), the patterns are quite similar, but the moisture values seemed smaller in the case of FLEX-WRF (Figs. 4k–n). The best results compared to the control was obtained with FLEX-ERA5.05 with an R value of ∼0.73 and the lowest STD of ∼0.46 mm day−1 (Fig. 5c). Although FLEX-ERA5.1 presented an R value of ∼0.7, a higher STD conditioned it than FLEX-ERA5.05, unlike FLEX-WRF with an STD < 0.2 mm day−1 but with a relatively lower R (∼0.65) than the other configurations. The B showed similar behavior as in winter, indicating a marked decrease for FLEX-ERA5.05 and FLEX-WRF but slightly lower for FLEX-ERA5.1. As seen in Figs. 6a and 6b, the maximum seasonal MAE and RMSE occurred in this season for all the configurations. FLEX-WRF showed the lowest MAE with ∼0.2 mm day−1, while FLEX-ERA5.05 and FLEX-ERA5.1 had higher values (∼1.99 and ∼1.88 mm day−1, respectively). On the other hand, RMSE showed similar values, with the lowest achieved for FLEX-ERA5.05 (∼0.48 mm day−1) and slightly higher values for FLEX-WRF and FLEX-ERA5.1 (∼0.5 mm day−1).

In autumn (OND), the values of the pattern showed differences over the IP (recycling processes), mainly for FLEX-ERA5.1 (Figs. 4p–s). Regarding the statistics, the lowest STD values were for FLEX-ERA5.05 and FLEX-WRF, showing similar values of ∼0.1 mm day−1; FLEX-ERA5.1 was slightly higher with ∼0.26 mm day−1. The B always showed a decrease, reaching its maximum for FLEX-ERA5.05; a pattern for which a slightly smoother field can be observed compared to the control configuration of FLEX-ERA-I.1. The lowest MAE was found for FLEX-ERA5.1 (∼0.16 mm day−1), although very similar to the other configurations. Regarding the RMSE, the values are the same for FLEX-WRF and FLEX-ERA5.1 (∼0.3 mm day−1) and slightly higher for FLEX-ERA5.05.

Annually the pattern was well represented for all the experiments (Figs. 4u–x), as shown in the correlation values of 0.7–0.8 for all configurations (Fig. 5e) and the low STD values for FLEX-ERA5.05 and FLEX-WRF (<0.2 mm day−1). The B was in the range of 0.02–0.06 mm day−1, showing a slight decrease compared to the control pattern. The MAE was around 0.1 mm day−1 for all configurations, but the RMSE showed the lowest value for FLEX-ERA5.05 (∼0.21 mm day−1), higher for FLEX-WRF (∼0.23 mm day−1) and FLEX-ERA5.1 (∼0.25 mm day−1).

In addition to the described results, the behavior of the models was also evaluated for some specific days of moisture transport to ensure that the fields were correctly represented in individual days (d). Figures S6S9 show the (EP)[d] > 0 values for days 1, 3, 5, and 10 backward in time, and Figs. S10S13 show the Taylor diagrams. In general, the R values ranged from values between 0.6 and 0.7 on day 1 backward, reaching ∼0.8 in some season values, with a decrease from day 1 to day 10 when R showed a range of 0.2–0.4. Decreases predominated for all configurations, showing a higher dispersion of values for FLEX-ERA5.1. In terms of correlations, FLEX-WRF has lower values. Finally, the MAE tends to have higher values the longer the time backward, especially for FLEX-ERA5.05 and FLEX-WRF showing maxima on days 3 and 5; however, the RSME shows its maximum on day 1, decreasing toward day 10 (Fig. S14).

Another test was done to check the ability of each FLEXPART configuration to adequately represent (EP)[d] > 0 values over the two areas defined by Gimeno et al. (2010b) as main moisture sources for the IP: the tropical-subtropical North Atlantic (TSNA) and the Mediterranean basin (IPM) (see Fig. S35). The temporal evolution of the contribution from both sources was quantified as in Nieto et al. (2006) and Gimeno et al. (2010a). Figure S15 shows the time series of (EP)[d] > 0 values over TSNA and IPM accounted for FLEX-ERA5.05 (red line), FLEX-ERA5.1 (blue line), FLEX-WRF (green line), and FLEX-ERA-I.1 (orange line). For winter and both sources (Figs. S33a,b), all configurations followed the control behavior, the maximum values were accounted for around days 4–7, and the minimum appeared on day 1. However, for TSNA the lowest contributions were obtained for FLEX-WRF and FLEX-ERA5.05 from day 2 to 10, and the highest for FLEX-ERA5.1. Between days 1 and 2, there was a slight increase for FLEX-WRF. For the IPM source from day 1 to 3, FLEX-ERA5.05 decreased the values. For spring, the TSNA series showed very similar values from days 7 to 1, reaching a higher convergence between the three configurations for day 4. For IPM all configurations reached their maximum for day 1, showing the highest decrease for FLEX-ERA5.05 (Figs. S33c,d). In summer the configurations behaved very similar for TSNA but differed for IPM, where FLEX-ERA5.05 and FLEX-WRF showed a considerable increase in day 1 compared with the maximum reached in the control configuration (Figs. S33e,f). In autumn for TSNA the furthest configuration from the control was FLEX-WRF, with a marked decrease up to day 4, while for IPM the values were well represented by most configurations (Figs. S33g,h). Overall, for the annual time series (Figs. S33j,k), the behavior of all configurations follows the control. The configuration that departs least from the TSNA was FLEX-ERA5.05 (followed by FLEX-ERA5.1 and FLEX-WRF) and FLEX-ERA5.1 for IPM (but with similar results for FLEX-WRF and FLEX-ERA5.05, being slightly better for the first one).

c. Representation of the moisture contribution for precipitation from the MED and the NATL sources

In this section, the aim is to test the ability of the three configurations to represent the sinks over the continents for the moisture coming from the MED and NATL sources. The (PE)[1–10] > 0 patterns are plotted in Fig. 7 for MED and Fig. 8 for NATL, showing the moisture sinks associated with both sources. The MED unequivocally influences the moisture budget in its surrounding continental area and, through the dominant local flows for the transport of air masses (Peixoto et al. 1982; Ward 1998; Nieto et al. 2010). The sinks identified by the control configuration FLEX-ERA-I.1 (Figs. 7d,i,n,s,x) over southern Europe, the Italian Peninsula, east of the IP, and north of the African continent [as in Gimeno et al. (2010b) or Ciric et al. (2016)] are congruent with the convergence pattern of the integrated vertical moisture flux (Figs. 7e,j,o,t,y), with maximum values over most of the areas mentioned above.

Fig. 7.
Fig. 7.

Moisture sink patterns (PE)[1–10] > 0; mm day−1) for the MED region from the Lagrangian outputs of FLEX-ERA5.05, FLEX-ERA5.1, FLEX-WRF, and FLEX-ERA-I.1. The right column represents the IVT field (arrows; kg m−1 s−1) and its divergence (contours; mm day−1) from ERA-I at 1°. The fields displayed correspond to (a)–(e) JFM, (f)–(j) AMJ, (k)–(o) JAS, (p)–(t) OND, and (u)–(y) annual.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

Fig. 8.
Fig. 8.

Taylor diagrams to compare the (PE)[1–10] > 0 fields of FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF with respect to the control field FLEX-ERA-I for the moisture sinks for the MED in terms of B, correlation coefficient, and STD. (a) JFM, (b) AMJ, (c) JAS, (d) OND, and (e) annual. The letters A, B, and C correspond to the configurations FLEX-ERA5.05, FLEX-ERA5.1, and FLEX-WRF, respectively.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

In winter, the three configurations showed high correlations with the control (0.7–0.8), and the STD ranged between 0.5 and 1.3 mm day−1 (Fig. 8a). There is a decrease for all the configurations, with lower values for FLEX-ERA5.1 (∼0.04 mm day−1) followed by FLEX-WRF (∼0.09 mm day−1) and FLEX-ERA5.05 (0.25 mm day−1). The MAE values were very similar to the other configurations, with a slightly lower error for FLEX-ERA5.1 (∼0.39 mm day−1, Fig. 6c) and the RMSE showing worse behavior. FLEX-WRF showed the lowest RMSE (∼0.82 mm day−1) followed by FLEX-ERA5.05 (∼0.87 mm day−1), as shown in Fig. 6d. The higher values of RSME for FLEX-ERA5.1 could be induced by the differences observed over the Balkan region (Fig. 7b).

In spring, the correlation decreased with respect to winter (0.6–0.7). The B (Fig. 8b) showed the smallest decrease for FLEX-WRF (∼0.02 mm day−1), followed by FLEX-ERA5.1 (∼0.06 mm day−1) and FLEX-ERA5.05 (∼0.10 mm day−1). STD showed a greater dispersion for FLEX-WRF and FLEX-ERA5.1 with values in the range of 1–1.5 mm day−1. There was a slight increase in error behavior compared to winter. The maximum values for the MAE were founded for FLEX-WRF, followed by FLEX-ERA5.1 and FLEX-ERA5.05 (see Fig. 6c) but with values equal to or lower than 0.50 mm day−1. Regarding the RMSE values, it was observed that all are similar, with the highest values for FLEX-ERA5.1 (∼1.11 mm day−1), followed by FLEX-WRF (∼1.09 mm day−1) and FLEX-ERA5.05 (∼1.00 mm day−1).

During summer, FLEX-ERA-I.1 showed a decrease in the moisture contribution associated with MED (Fig. 8n), and the configurations show the highest STDs values, in the range of 0.7–1.8 mm day−1, and the lowest R values (∼0.6) of all seasons. FLEX-WRF presents a slight increase of ∼0.03 mm day−1 with respect to the control, in contrast to the decrease observed of FLEX-ERA5.1 and FLEX-ERA5.05 (Fig. 8c). The three configurations showed the highest MAE and RMSE compared with the other seasons, with values from 1.30 to 1.48 mm day−1 (Figs. 6c,d).

For autumn, the correlation ranged from approximately 0.7–0.8, superior to previous seasons, but higher STD values were also reached, with approximately 0.92–1.52 mm day−1, only surpassed by the summer season. All configurations decreased the precipitation over continental areas, but the lower B was achieved for FLEX-WRF (∼0.02 mm day−1) with similar values for FLEX-ERA5.1 ∼ 0.03 mm day−1). The MAE decreased for FLEX-ERA5.1 and FLEX-ERA5.05 but had a similar behavior for FLEX-WRF compared to summer. The RMSE also decreased concerning the previous period, with the best results for FLEX-ERA5.05 (∼1.1 mm day−1) (Fig. 6d).

Finally, the correlations for the annual pattern showed values from 0.7 to 0.8, with a lower decrease for FLEX-WRF (Fig. 8e). The MAE was lower for FLEX-ERA5.05 and slightly higher with similar values for FLEX-ERA5.1 and FLEX-WRF; however, the highest RMSE was achieved by FLEX-ERA5.1

The patterns for individual days (PE)[d] > 0 (for d = 1, 3, 5, and 10) are plotted in Figs. S15–S18; the Taylor diagrams for these days are shown in Figs. S19–S22. In general, R values were higher at day 1 in a range of 0.6–0.7, reaching values in certain seasons up to approximately 0.85, and with a considerable decrease to day 10, with values between 0.2 and 0.5. Overall, there were decreases for all configurations, although there was a marked increase for FLEX-WRF on day 1. The highest dispersion for the sink field was observed for FLEX-ERA5.1, and the lower correlation occurred for FLEX-WRF. The MAE followed a similar behavior for the analyzed days, with day 3 standing out for its high values for FLEX-ERA5.05 and FLEX-WRF. The RSME remained very similar for the three configurations for all seasons (see Fig. S23).

Analogous to the MED analysis, the sink patterns for the NATL moisture source are shown in Fig. 9. In winter, the three configurations showed similar contributions for precipitation compared to the control configuration. However, some differences can be observed, such as the smoother pattern for FLEX-ERA5.05 (Fig. 9a) or the more intense values for FLEX-ERA5.1 (Fig. 9b). Compared with the remaining seasons, winter achieved the maximum correlation values (∼0.7–0.8) for the three configurations versus the control one, with FLEX-ERA5.05 at the top. The STD ranged around 2–4 mm day−1, showing that FLEX-ERA5.1 had a greater dispersion and that FLEX-WRF had the lowest (with a R ∼ 0.7). Although FLEX-WRF and FLEX-ERA5.05 tended to decrease the pattern (with 0.66 and 0.78 mm day−1, respectively), FLEX-ERA5.1 increased it (0.04 mm day−1), as shown in Fig. 10a. On the other hand (see Figs. 6e,f), this last configuration presented the lowest MAE (∼1.36 mm day−1), followed by FLEX-ERA5.05 and FLEX-WRF (with values of 1.48 and 1.60 mm day−1, respectively). FLEX-ERA5.05 had the lowest the RMSE (∼2.52 mm day−1) followed by FLEX-ERA5.1 and FLEX-WRF (with values of 2.67 and 2.74 mm day−1, respectively).

Fig. 9.
Fig. 9.

As in Fig. 7, but for the moisture sinks from the NATL source.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

Fig. 10.
Fig. 10.

Taylor diagrams as in Fig. 8, but for the moisture sinks from the NATL.

Citation: Journal of Hydrometeorology 24, 2; 10.1175/JHM-D-22-0018.1

In spring, there was a decrease in the contribution pattern and in its intensity (Fig. 9i). The correlation values were lower than in winter reaching the maximum values for FLEX-ERA5.05 and FLEX-ERA5.1, with ∼0.6 and ∼0.5, respectively. The STD presents values in the range of 2–4 mm day−1, showing a decrease of the pattern with the highest value of B (absolute value) for FLEX-ERA5.05 (∼1 mm day−1), followed by FLEX-WRF (∼0.6 mm day−1) and finally FLEX-ERA5.1 (∼0.2 mm day−1) (see Fig. 10b). The MAE showed the same values of ∼1.5 mm day−1 for FLEX-ERA5.05 and FLEX-ERA5.1, and 1.68 mm day−1 for FLEX-WRF (see Fig. 6e). The RMSE showed that more marked differences were maintained in certain mesh points for FLEX-ERA5.1 represented by values of ∼3 mm day−1, followed by FLEX-WRF with ∼2.88 mm day−1, and FLEX-ERA5.05 ∼ 2.78 mm day−1.

For the summer, a very weak pattern of (PE)[1–10] > 0 continued with a very dispersed structure, as shown in Figs. 9k–n, with maximums located at low latitudes. The contribution pattern is located very close to the east coast of the United States and Canada, influencing the islands of the Caribbean and Central America at more than 20 mm day−1. In relation to the statistics, the FLEX-WRF statistics presents an R ∼ 0.4; for the rest of the configurations, it was ∼0.6. The STD shows very similar values for FLEX-ERA5.05 and FLEX-WRF but higher values for FLEX-ERA5.1 at ∼4.68 mm day−1. On the other hand, B showed decreases for FLEX-ERA5.05 and FLEX-ERA5.1 with values ∼ 1 and ∼0.15 mm day−1, respectively; these were different from FLEX-WRF, which increased the pattern by 0.32 mm day−1 (see Fig. 10c). Furthermore, FLEX-ERA5.1 shows the maximum RMSE for the entire time interval at ∼3.56 mm day−1 (see Fig. 6f).

In autumn, the STD decreases with respect to the previous one, with only FLEX-ERA5.1 exceeding 4 mm day−1. For all simulations, the decrease with the highest value corresponded to FLEX-ERA5.05 with ∼1.14 mm day−1. The highest value of MAE was obtained for FLEX-ERA5.05 (∼2.3 mm day−1), while the RMSE for FLEX-WRF and FLEX-ERA5.1 reached ∼3.8 mm day−1 (see Fig. 10d).

For the annual period Fig. 10e shows that the correlation for all configurations was high with values in the 0.7–0.8 range, and with an STD < 2 mm day−1 for FLEX-WRF and FLEX-ERA5.05; this was relatively higher for FLEX-ERA5.1. The B showed that FLEX-ERA5.1 presented a slight increase; for the remaining ones there were a lower decrease, with values of ∼0.2 mm day−1 for FLEX-WRF. The MAE showed a similar behavior close to 1 mm day−1 for FLEX-ERA5.05 and FLEX-ERA5.1 but slightly higher for FLEX-WRF (see Fig. 6e). The RMSE values were <2 mm day−1 for FLEX-WRF and FLEX-ERA5.05 but >2 mm day−1 for FLEX-ERA5.1, showing that it presents difficulties in representing the sink pattern for certain regions.

The representation of the seasonal and annual sink patterns for day 1, 3, 5, and 10 are shown in Figs. S24–S26. Overall, by seasons the B showed differences between the configurations with some tendency to increase in the first days for FLEX-WRF and FLEX-ERA5.1. Correlation values (R) were more stable for NATL than MED, ranging ∼0.4–0.6, reaching slightly lower values for FLEX-WRF. The three configurations show similarities to the case of MED in terms of dispersion, increasing the STD from day 1 to day 10 (see Figs. S28–S31). Finally, all configurations show a tendency to increase MAE and RSME from day 1 to day 10 with maxima for FLEX-WRF and FLEX-ERA5.1, respectively. Annually, it was observed that FLEX-ERA5.05 and FLEX-WRF tend to have slightly higher MAE values than FLEX-ERA5.1, while the latter configuration predominantly has higher RMSE values than the other two configurations (Figs. S14, S23, and S32). The Taylor diagrams for the individual days (Fig. S34) show that the R decreases as the days increased backward. FLEX-WRF presents the lower R in most cases, while the highest dispersion is associated with FLEX-ERA5.1. For FLEX-ERA5.05, there is a tendency to decrease the positive or negative EP values in all cases.

4. Discussions and uncertainty analysis

The results obtained for the analysis of the moisture sources that contribute to the IP show that, in the four seasons and the annual period, the correlation for the moisture sources pattern with respect to the FLEX-ERA-I.1 (the control simulation) for the three configurations analyzed varied from 0.4 to 0.6, higher for FLEX-ERA5.05 and lower for FLEX-WRF. In addition, there were few marked differences between the configurations in terms of MAE and RMSE, and according to these statistics the lowest performance was observed for summer and the best for winter and autumn. These results were corroborated by R2 values, being the maximum for FLEX-ERA5.05 in most of the studied periods, ranging from 0.25 (winter) to 0.55 (annual period). However, the reverse occurs for the VAR, reaching the minimum for winter and the maximum for the annual period. This is due to the lower mean moisture values for the annual period, which increases the value of the VAR (see Table S1). According to these statistics, FLEX-ERA5.1 and FLEX-WRF showed very similar behavior for R2, although the VAR was slightly lower for FLEX-WRF (Table S1).

On the other hand, the FLEX-WRF configuration indicates some difficulty in representing the source above 40°N over the North Atlantic Ocean, areas with significant differences (at 99%) are observed with respect to the control experiment (Fig. S38). This fact could be due to the regional characteristics of the model. Therefore, a certain number of particles disappear across the boundaries (see Fig. 1a) in their movement, limiting their contribution to the moisture budget balance. This could also influence the statistics that characterize the comparative study between the different set of configurations for the analysis of the moisture source patterns. However, FLEX-WRF is the configuration that represents better the pattern for the moisture sources in regions with complex orography, especially during spring and summer, the periods when the precipitation is due to recycling processes (see Fig. S36). This shows that the use of the WRF-ARW allows a detailed representation of the most local processes of moisture transport (e.g., on the IP). This could be associated with the surface parameterization of WRF-ARW using the Noah LSM model that allows a better representation of the sensible and latent heat fluxes provided to the boundary layer scheme and of the longwave and shortwave radiation reflected, variables involved in moisture transport from the surface to different levels of the atmosphere.

For the analysis of moisture sinks for the MED region, the three configurations compared showed a low dispersion in their fields compared with the control. This is shown by the high values of R ≥ 0.6. In addition, R2 values ranged from 0.3 to 0.6, showing a good correlation, being higher in all periods for FLEX-ERA5.05, and the VAR was lower for FLEX-WRF (1.6–2.37) followed by FLEX-ERA5.05 (1.89–2.32) (Table S2). This behavior for FLEX-ERA5.05 may be associated with the fact that over the moisture sinks a larger number of areas with significant differences are observed (Fig. S39), mainly in the mountainous regions of northern Italy and the Balkan Peninsula. Also, for FLEX-WRF it is important to note that the most significant differences founded around 30°W (Fig. S39) could be related to the closeness of the boundary of the FLEXPART-WRF domain, increasing possible errors due to the possible crossing of some trajectories through it. In addition, a lower MAE was achieved although the order of magnitude of the moisture budget for the MED was higher. The RMSE was considerable, showing an increase in summer and then a decrease in autumn. On the other hand, similar to the results for the IP, FLEX-WRF showed the best behavior for the moisture sink pattern in mountainous regions such as the Alps, the Balkan Peninsula, and eastern Europe (Fig. S37).

Finally, in general, the contribution pattern of moisture sinks from the NATL tends to be larger in latitudinal and longitudinal dimension, causing the errors to be more noticeable at each season compared to those obtained for MED. The correlation between the three configurations and FLEX-ERA-I.1 ranged ∼0.4–0.7; however, in this case, there is an increase in the magnitude of MAE and RMSE, ranging from 1 to 2.5 mm day−1 and from 2 to 4.5 mm day−1, respectively. For this target region, the lowest values for the coefficient of variation were from 0.9 to 1.5, being better than for the other target regions analyzed. Regarding the R2, FLEX-ERA5.05 showed values from 0.3 to 0.6, followed by FLEX-ERA5.1, and with lower values for FLEX-WRF (Table S3). In this case, it is observed that FLEX-ERA5.1 shows the least number of points with significant differences (at 99%) compared to FLEX-ERA-I.1, although the values of the sink pattern are higher for the entire North Atlantic region. On the other hand, FLEX-ERA5.05 shows a behavior with small differences for the spring, summer, and autumn, but for winter and annual scale some areas showed significant differences due to regions with lower values compared to the control experiment (Fig. S40). FLEX-WRF shows difficulty in representing the pattern, with very low values in the region of maximum values shown by FLEX-ERA-I.1. For this oceanic region FLEX-WRF could present some limitations and therefore the use of FLEX-ERA5.05 or FLEX-ERA5.1 would be more efficient.

The behavior of the errors obtained in the comparison of the configurations with respect to ERA-I shows that there are few differences for the patterns of moisture sources and sinks with the use of different versions of reanalysis, similar to that obtained by Durán-Quesada et al. (2010, 2017).

5. Summary and conclusions

Here, we compared the representation of moisture sources and sinks for three different simulations of the Lagrangian model FLEXPART, using the latest FLEXPART model version (v10.3) with the new ERA5 reanalysis as input data at two horizontal resolutions (1° and 0.5°), and the FLEXPART-WRF version fed with the dynamic downscaling WRF-ARW outputs (at a 0.18° horizontal resolution) previously forced with ERA5. The simulations were done over the extended area of the North Atlantic and European region, because it comprises two of the main oceanic global moisture sources, the so-called NATL (North Atlantic ocean) and MED (the whole Mediterranean Sea). So, the target regions analyzed were the moisture sinks for these two oceanic sources, and the moisture sources for the Iberian Peninsula (IP). The year selected for analysis was a year with neutral characteristics in terms of the climate variability patterns affecting the region, which is 2014. The outputs of the previous FLEXPARTv9.0 model version forced with ERA-I (at 1°) were used as control experiments.

An analysis of variables related to the water cycle (such as total precipitation, evaporation, total column water, integrated water vapor transport, specific humidity at 850 hPa) showed that there are few significant differences (at 95%) between the ERA5 and ERA-I reanalyses. Nogueira (2020) affirms that this is enough for an improved representation of moisture sink/source patterns. This improvement in ERA5 versus ERA-I is crucial for our study of moisture transport, because in addition some of the regions with significant improvements are located in our study region, acting as sources of moisture.

Annually, the FLEX-ERA5.05 configuration showed the most similar representation compared to the control experiment. Statistics show the maximum correlation and the lowest error values compared to the FLEX-ERA-I.1 simulations, albeit with a tendency to decrease the values of the moisture budget. The FLEX-WRF configuration was the second in representing sources and sinks, showing a slight increase for MED. Finally, FLEX-ERA5.1 had the lowest representation, with the maximum errors and maximum deviation for the three target regions.

For all configurations, winter is when the three simulations showed the greatest similarity with respect to the control one, indicating discrepancies higher in summer and autumn. In general, the best results for all configurations were obtained for the sinks associated with the MED moisture source, with maximum correlation values reached for all seasons and the lowest MAE compared to the control experiment.

On the other hand, it should be noted that FLEX-WRF is the configuration that represents better the pattern of moisture source or sink in regions with complex orography and for recycling processes over the IP or in the mountainous regions around the Mediterranean Sea. This shows that the use of WRF-ARW allows a detailed representation of the most local processes of moisture transport.

The analysis per day shows greater differences for each configuration compared to the integrated pattern along the 10 days of the trajectories; this may be associated with a smoothing of the pattern when adding all the values (positive and negative) from day 1 to 10.

The main differences of FLEX-ERA5.1 versus FLEX-ERA5.05 are related to the number of particles modeled. The global run experiment for these configurations use 9 and 30 million particles, respectively, generating readjustments in the moisture content distribution due to different vertical and horizontal resolutions compared to the 2 million particles used for FLEX-ERA-I.1, but with a higher standard deviation for FLEX-ERA5.1 and a smoother pattern for FLEX-ERA5.05.

In summary, climatological studies using FLEXPARTv10.3 forced with ERA5 reanalysis data at various horizontal resolutions (0.5° and 1°) represent well the position and extension of moisture sources and sink patterns when forced by ERA-Interim (1°), as well as the amounts of moisture modeled. In addition, the FLEXPART version fed with the dynamic downscaled WRF-ARW outputs at a higher resolution (∼0.25°, about ∼20 km), previously forced with ERA5, accurately represents these patterns but with more detail, in particular in orographic regions. This improvement even in climatological studies, such as this one, in which the fields are smoothed through the seasonal or annual averages carried out, indicates that for shorter periods, specific events or case studies (e.g., tropical or extratropical cyclones) the smaller scale supported by the WRF-ARW data would help to determine transport patterns or structures in more detail.

Acknowledgments.

J.C.F.-A. and M.V. acknowledge the support from the Xunta de Galicia (Galician Regional Government) under Grants ED481A-2020/193 and ED481B 2018/062, respectively. A.P.-A. acknowledges a PhD grant from the University of Vigo. This work is supported by the SETESTRELO project (PID2021-122314OB-I00) funded by the Ministerio de Ciencia, Innovación y Universidades, Spain. Partial support was also obtained from the Xunta de Galicia under the project “Programa de Consolidación e Estructuración de Unidades de Investigación Competitivas (Grupos de Referencia Competitiva)” (ED431C 2021/44). In addition, this work has been possible thanks to the computing resources and technical support provided by CESGA (Centro de Supercomputación de Galicia) and Red Española de Supercomputación (RES) (AECT-2022-3-0009 and DATA-2021-1-0005).

Data availability statement.

ERA5 reanalysis data can be obtained from https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels-monthly-means?tab=form and ERA-Interim data from https://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/. The FLEXPART and FLEXPART-WRF outputs are available upon request to the corresponding author.

REFERENCES

  • Bechtold, P., M. Köhler, T. Jung, F. Doblas-Reyes, M. Leutbecher, M. J. Rodwell, F. Vitart, and G. Balsamo, 2008: Advances in simulating atmospheric variability with the ECMWF model: From synoptic to decadal time-scales. Quart. J. Roy. Meteor. Soc., 134, 13371351, https://doi.org/10.1002/qj.289.

    • Search Google Scholar
    • Export Citation
  • Berkowicz, R., and L. P. Prahm, 1982: Sensible heat flux estimated from routine meteorological data by the resistance method. J. Appl. Meteor., 21, 18451864, https://doi.org/10.1175/1520-0450(1982)021%3C1845:SHFEFR%3E2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brioude, J., and Coauthors, 2013: The Lagrangian particle dispersion model FLEXPART-WRF version 3.1. Geosci. Model Dev., 6, 18891904, https://doi.org/10.5194/gmd-6-1889-2013.

    • Search Google Scholar
    • Export Citation
  • Brown, B., E. Ebert, T. Fowler, E. Gilleland, P. Kucera, and L. Wilson, 2013: Verification methods for tropical cyclone forecasts. WMO Tech. Doc. WWRP 2013-7, 89 pp.

  • Castillo, R., R. Nieto, A. Drumond, and L. Gimeno, 2014: Estimating the temporal domain when the discount of the net evaporation term affects the resulting net precipitation pattern in the moisture budget using a 3-D Lagrangian approach. PLOS ONE, 9, e99046, https://doi.org/10.1371/journal.pone.0099046.

    • Search Google Scholar
    • Export Citation
  • Ciric, D., M. Stojanovic, A. Drumond, R. Nieto, and L. Gimeno, 2016: Tracking the origin of moisture over the Danube River Basin using a Lagrangian approach. Atmosphere, 12, 162, https://doi.org/10.3390/atmos7120162.

    • Search Google Scholar
    • Export Citation
  • Cloux, S., D. Garaboa-Paz, D. Insua-Costa, G. Miguez-Macho, and V. Pérez-Muñuzuri, 2021: Extreme precipitation events in the Mediterranean area: Contrasting Lagrangian and Eulerian models for moisture sources identification. Hydrol. Earth Syst. Sci., 25, 64656477, https://doi.org/10.5194/hess-25-6465-2021.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, https://doi.org/10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, L. Gimeno, and T. Ambrizzi, 2008: A Lagrangian identification of major sources of moisture over central Brazil and La Plata Basin. J. Geophys. Res., 113, D14128, https://doi.org/10.1029/2007JD009547.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, and L. Gimeno, 2011a: On the contribution of the Tropical Western Hemisphere Warm Pool source of moisture to the Northern Hemisphere precipitation through a Lagrangian approach. J. Geophys. Res., 116, D00Q04, https://doi.org/10.1029/2010JD015397.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, E. Hernandez, and L. Gimeno, 2011b: A Lagrangian analysis of the variation in moisture sources related to drier and wetter conditions in regions around the Mediterranean Basin. Nat. Hazards Earth Syst. Sci., 11, 23072320, https://doi.org/10.5194/nhess-11-2307-2011.

    • Search Google Scholar
    • Export Citation
  • Durán-Quesada, A. M., L. Gimeno, J. A. Amador, and R. Nieto, 2010: Moisture sources for Central America: Identification of moisture sources using a Lagrangian analysis technique. J. Geophys. Res., 115, D05103, https://doi.org/10.1029/2009JD012455.

    • Search Google Scholar
    • Export Citation
  • Durán-Quesada, A. M., L. Gimeno, and J. Amador, 2017: Role of moisture transport for Central American precipitation. Earth Syst. Dyn., 8, 147161, https://doi.org/10.5194/esd-8-147-2017.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., and M. Živković-Rothman, 1999: Development and evaluation of a convection scheme for use in climate models. J. Atmos. Sci., 56, 17661782, https://doi.org/10.1175/1520-0469(1999)056%3C1766:DAEOAC%3E2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., R. Nieto, R. Trigo, S. M. Vicente-Serrano, and J. I. López-Moreno, 2010a: Where does the Iberian Peninsula moisture come from? An answer based on a Lagrangian approach. J. Hydrometeor., 11, 421436, https://doi.org/10.1175/2009JHM1182.1.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., A. Drumond, R. Nieto, R. M. Trigo, and A. Stohl, 2010b: On the origin of continental precipitation. Geophys. Res. Lett., 37, L13804, https://doi.org/10.1029/2010GL043712.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., and Coauthors, 2012: Oceanic and terrestrial sources of continental precipitation. Rev. Geophys., 50, RG4003, https://doi.org/10.1029/2012RG000389.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., and Coauthors, 2020: Recent progress on the sources of continental precipitation as revealed by moisture transport analysis. Earth-Sci. Rev., 201, 103070, https://doi.org/10.1016/j.earscirev.2019.103070.

    • Search Google Scholar
    • Export Citation
  • Gómez-Hernández, M., A. Drumond, L. Gimeno, and R. Garcia-Herrera, 2013: Variability of moisture sources in the Mediterranean region during the period 1980-2000. Water Resour. Res., 49, 67816794, https://doi.org/10.1002/wrcr.20538.

    • Search Google Scholar
    • Export Citation
  • Grythe, H., N. I. Kristiansen, C. D. Groot-Zwaaftink, S. Eckhardt, P. Tunved, R. Krejci, J. Ström, and A. Stohl, 2017: A new aerosol wet removal scheme for the Lagrangian particle model FLEXPART v10. Geosci. Model Dev., 10, 14471466, https://doi.org/10.5194/gmd-10-1447-2017.

    • Search Google Scholar
    • Export Citation
  • Hanna, S. R., 1982: Applications in air pollution modeling. Atmospheric Turbulence and Air Pollution Modelling, F. T. M. Nieuwstadt and H. Van Dop, Eds., D. Reidel Publishing Company, 275–310.

  • Hersbach, H., and D. Dee, 2016: ERA5 reanalysis is in production. ECMWF Newsletter, No. 147, ECMWF, Reading, United Kingdom, 7, http://www.ecmwf.int/sites/default/files/elibrary/2016/16299-newsletter-no147-spring-2016.pdf.

  • Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 19992049, https://doi.org/10.1002/qj.3803.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., and J. O. J. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129151.

  • Hong, S. Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341, https://doi.org/10.1175/MWR3199.1.

    • Search Google Scholar
    • Export Citation
  • Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long–lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.

    • Search Google Scholar
    • Export Citation
  • Insua-Costa, D., and G. Miguez-Macho, 2018: A new moisture tagging capability in the Weather Research and Forecasting model: Formulation, validation and application to the 2014 Great Lake-effect snowstorm. Earth Syst. Dyn., 9, 167185, https://doi.org/10.5194/esd-9-167-2018.

    • Search Google Scholar
    • Export Citation
  • Insua-Costa, D., G. Miguez-Macho, and M. C. Llasat, 2019: Local and remote moisture sources for extreme precipitation: A study of the two catastrophic 1982 western Mediterranean episodes. Nat. Hazards Earth Syst. Sci., 23, 38853900, https://doi.org/10.5194/hess-23-3885-2019.

    • Search Google Scholar
    • Export Citation
  • Jimenez, P. A., J. Dudhia, J. F. Gonzalez-Rouco, J. Navarro, J. P. Montavez, and E. Garcia–Bustamante, 2012: A revised scheme for the WRF surface layer formulation. Mon. Wea. Rev., 140, 898918, https://doi.org/10.1175/MWR-D-11-00056.1.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NMC/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lavers, D. A., and G. Villarini, 2013: The nexus between atmospheric rivers and extreme precipitation across Europe. Geophys. Res. Lett., 40, 32593264, https://doi.org/10.1002/grl.50636.

    • Search Google Scholar
    • Export Citation
  • Lavers, D. A., G. Villarini, R. P. Allan, E. F. Wood, and A. J. Wade, 2012: The detection of atmospheric rivers in atmospheric reanalyses and their links to British winter floods and the large-scale climatic circulation. J. Geophys. Res., 117, D20106, https://doi.org/10.1029/2012JD018027.

    • Search Google Scholar
    • Export Citation
  • Lehner, B., K. Verdin, and A. Jarvis, 2008: New global hydrography derived from spaceborne elevation data. Eos, Trans. Amer. Geophys. Union, 89, 9394, https://doi.org/10.1029/2008EO100001.

    • Search Google Scholar
    • Export Citation
  • Lionello, P., and Coauthors, 2006: The Mediterranean climate: An overview of the main characteristics and issues. Dev. Earth Environ. Sci., 4, 126, https://doi.org/10.1016/S1571-9197(06)80003-0.

    • Search Google Scholar
    • Export Citation
  • Miguez-Macho, G., G. L. Stenchikov, and A. Robock, 2004: Spectral nudging to eliminate the effects of domain position and geometry in regional climate model simulations. J. Geophys. Res., 109, D13104, https://doi.org/10.1029/2003JD004495.

    • Search Google Scholar
    • Export Citation
  • Nieto, R., L. Gimeno, and R. M. Trigo, 2006: A Lagrangian identification of major sources of Sahel moisture. Geophys. Res. Lett., 33, L18707, https://doi.org/10.1029/2006GL027232.

    • Search Google Scholar
    • Export Citation
  • Nieto, R., L. Gimeno, A. Drumond, and E. Hernandez, 2010: A Lagrangian identification of the main moisture sources and sinks affecting the Mediterranean area. WSEAS Trans. Environ. Dev., 6, 365374.

    • Search Google Scholar
    • Export Citation
  • Nogueira, M., 2020: Inter-comparison of ERA-5, ERA-Interim and GPCP rainfall over the last 40 years: Process-based analysis of systematic and random differences. J. Hydrol., 583, 124632, https://doi.org/10.1016/j.jhydrol.2020.124632.

    • Search Google Scholar
    • Export Citation
  • Numaguti, A., 1999: Origin and recycling processes of precipitating water over the Eurasian continent: Experiments using an atmospheric general circulation model. J. Geophys. Res., 104, 19571972, https://doi.org/10.1029/1998JD200026.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., M. De Almeida, R. D. Rosen, and D. A. Salstein, 1982: Atmospheric moisture transport and the water balance of the Mediterranean Sea. Water Resour. Res., 18, 8390, https://doi.org/10.1029/WR018i001p00083.

    • Search Google Scholar
    • Export Citation
  • Pisso, I., and Coauthors, 2019: The Lagrangian particle dispersion model FLEXPART version 10.4. Geosci. Model Dev., 12, 49554997, https://doi.org/10.5194/gmd-12-4955-2019.

    • Search Google Scholar
    • Export Citation
  • Quante, M. V., and V. Mathias, 2006: Water in the Earth’s atmosphere. J. Phys. IV France, 139, 3761, https://doi.org/10.1051/jp4:2006139005.

    • Search Google Scholar
    • Export Citation
  • Rahimi, S., W. Krantz, Y.-H. Lin, B. Bass, N. Goldenson, A. Hall, Z. J. Lebo, and J. Norris, 2022: Evaluation of a reanalysis‐driven configuration of WRF4 over the western United States from 1980 to 2020. J. Geophys. Res. Atmos., 127, e2021JD035699, https://doi.org/10.1029/2021JD035699.

    • Search Google Scholar
    • Export Citation
  • Ramos, A. M., M. J. Martins, R. Tomé, and R. M. Trigo, 2018: Extreme precipitation events in summer in the Iberian Peninsula and its relationship with atmospheric rivers. Front. Earth Sci., 6, 110, https://doi.org/10.3389/feart.2018.00110.

    • Search Google Scholar
    • Export Citation
  • Randhir, T. O., 2012: Water for life and ecosystem sustainability. Earth Sci. Climatic Change, 3, e107, https://doi.org/10.4172/2157-7617.1000e107.

    • Search Google Scholar
    • Export Citation
  • Rios-Entenza, A., and G. Miguez-Macho, 2014: Moisture recycling and the maximum of precipitation in spring in the Iberian Peninsula. Climate Dyn., 42, 32073231, https://doi.org/10.1007/s00382-013-1971-x.

    • Search Google Scholar
    • Export Citation
  • Rios-Entenza, A., P. M. M. Soares, R. M. Trigo, R. M. Cardoso, and G. Miguez-Macho, 2014: Moisture recycling in the Iberian Peninsula from a regional climate simulation: Spatiotemporal analysis and impact on the precipitation regime. J. Geophys. Res. Atoms., 119, 58955912, https://doi.org/10.1002/2013JD021274.

    • Search Google Scholar
    • Export Citation
  • Schicker, I., S. Radanovics, and P. Seibert, 2010: Origin and transport of Mediterranean moisture and air. Atmos. Chem. Phys., 10, 50895105, https://doi.org/10.5194/acp-10-5089-2010.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., http://dx.doi.org/10.5065/D68S4MVH.

  • Sodemann, H., and A. Stohl, 2013: Moisture origin and meridional transport in atmospheric rivers and their association with multiple cyclones. Mon. Wea. Rev., 141, 28502868, https://doi.org/10.1175/MWR-D-12-00256.1.

    • Search Google Scholar
    • Export Citation
  • Stohl, A., 1998: Computation, accuracy and applications of trajectories—A review and bibliography. Atmos. Environ., 32, 947966, https://doi.org/10.1016/S1352-2310(97)00457-3.

    • Search Google Scholar
    • Export Citation
  • Stohl, A., and D. J. Thomson, 1999: A density correction for Lagrangian particle dispersion models. Bound.-Layer Meteor., 90, 155167, https://doi.org/10.1023/A:1001741110696.

    • Search Google Scholar
    • Export Citation
  • Stohl, A., and P. A. James, 2004: Lagrangian analysis of the atmospheric branch of the global water cycle. Part I: Method description, validation, and demonstration for the August 2002 flooding in Central Europe. J. Hydrometeor., 5, 656678, https://doi.org/10.1175/1525-7541(2004)005<0656:ALAOTA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Stohl, A., and P. A. James, 2005: A Lagrangian analysis of the atmospheric branch of the global water cycle: Part II: Moisture transports between Earth’s ocean basins and river catchments. J. Hydrometeor., 6, 961984, https://doi.org/10.1175/JHM470.1.

    • Search Google Scholar
    • Export Citation
  • Stohl, A., C. Forster, A. Frank, P. Seibert, and G. Wotawa, 2005: Technical note: The Lagrangian particle dispersion model FLEXPART version 6.2. Atmos. Chem. Phys., 5, 24612474, https://doi.org/10.5194/acp-5-2461-2005.

    • Search Google Scholar
    • Export Citation
  • Tewari, M., and Coauthors, 2004: Implementation and verification of the Unified NOAH land surface model in the WRF Model. 20th Conf. on Weather Analysis and Forecasting/16th Conf. on Numerical Weather Prediction, Seattle, WA, Amer. Meteor. Soc., 14.2a, https://ams.confex.com/ams/84Annual/techprogram/paper_69061.htm.

  • Thomson, D. J., 1987: Criteria for the selection of stochastic models of particle trajectories in turbulent flows. J. Fluid Mech., 180, 529556, https://doi.org/10.1017/S0022112087001940.

    • Search Google Scholar
    • Export Citation
  • Trigo, I. F., T. D. Davies, and G. R. Bigg, 1999: Objective climatology of cyclones in the Mediterranean region. J. Climate, 12, 16851696, https://doi.org/10.1175/1520-0442(1999)012<1685:OCOCIT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Trigo, I. F., T. D. Davies, and G. R. Bigg, 2000: Decline in Mediterranean rainfall caused by weakening of Mediterranean cyclones. Geophys. Res. Lett., 27, 29132916, https://doi.org/10.1029/2000GL011526.

    • Search Google Scholar
    • Export Citation
  • van der Ent, R. J., and O. A. Tuinenburg, 2017: The residence time of water in the atmosphere revisited. Hydrol. Earth Syst. Sci., 21, 779790, https://doi.org/10.5194/hess-21-779-2017.

    • Search Google Scholar
    • Export Citation
  • Vázquez, M., R. Nieto, M. L. R. Liberato, and L. Gimeno, 2020: Atmospheric moisture sources associated with extreme precipitation during the peak precipitation month. Wea. Climate Extremes, 30, 100289, https://doi.org/10.1016/j.wace.2020.100289.

    • Search Google Scholar
    • Export Citation
  • Ward, M. N., 1998: Diagnosis and short-lead predictions of summer rainfall in tropical North Africa at interannual and multidecadal timescales. J. Climate, 11, 31673191, https://doi.org/10.1175/1520-0442(1998)011<3167:DASLTP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Warner, T. T., R. A. Peterson, and R. E. Peterson, 1997: A tutorial on lateral boundary conditions as a basic and potentially serious limitation to regional weather prediction. Bull. Amer. Meteor. Soc., 78, 25992617, https://doi.org/10.1175/1520-0477(1997)078<2599:ATOLBC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zhang, Z., F. M. Ralph, and M. Zheng, 2019: The relationship between extratropical cyclone strength and atmospheric river intensity and position. Geophys. Res. Lett., 46, 18141823, https://doi.org/10.1029/2018GL079071.

    • Search Google Scholar
    • Export Citation
  • Zhao, N., A. Manda, X. Guo, K. Kikuchi, T. Nasuno, M. Nakano, Y. Zhang, and B. Wang, 2021: A Lagrangian view of moisture transport related to the heavy rainfall of July 2020 in Japan: Importance of the moistening over the subtropical regions. Geophys. Res. Lett., 48, e2020GL091441, https://doi.org/10.1029/2020GL091441.

    • Search Google Scholar
    • Export Citation

Supplementary Materials

Save
  • Bechtold, P., M. Köhler, T. Jung, F. Doblas-Reyes, M. Leutbecher, M. J. Rodwell, F. Vitart, and G. Balsamo, 2008: Advances in simulating atmospheric variability with the ECMWF model: From synoptic to decadal time-scales. Quart. J. Roy. Meteor. Soc., 134, 13371351, https://doi.org/10.1002/qj.289.

    • Search Google Scholar
    • Export Citation
  • Berkowicz, R., and L. P. Prahm, 1982: Sensible heat flux estimated from routine meteorological data by the resistance method. J. Appl. Meteor., 21, 18451864, https://doi.org/10.1175/1520-0450(1982)021%3C1845:SHFEFR%3E2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brioude, J., and Coauthors, 2013: The Lagrangian particle dispersion model FLEXPART-WRF version 3.1. Geosci. Model Dev., 6, 18891904, https://doi.org/10.5194/gmd-6-1889-2013.

    • Search Google Scholar
    • Export Citation
  • Brown, B., E. Ebert, T. Fowler, E. Gilleland, P. Kucera, and L. Wilson, 2013: Verification methods for tropical cyclone forecasts. WMO Tech. Doc. WWRP 2013-7, 89 pp.

  • Castillo, R., R. Nieto, A. Drumond, and L. Gimeno, 2014: Estimating the temporal domain when the discount of the net evaporation term affects the resulting net precipitation pattern in the moisture budget using a 3-D Lagrangian approach. PLOS ONE, 9, e99046, https://doi.org/10.1371/journal.pone.0099046.

    • Search Google Scholar
    • Export Citation
  • Ciric, D., M. Stojanovic, A. Drumond, R. Nieto, and L. Gimeno, 2016: Tracking the origin of moisture over the Danube River Basin using a Lagrangian approach. Atmosphere, 12, 162, https://doi.org/10.3390/atmos7120162.

    • Search Google Scholar
    • Export Citation
  • Cloux, S., D. Garaboa-Paz, D. Insua-Costa, G. Miguez-Macho, and V. Pérez-Muñuzuri, 2021: Extreme precipitation events in the Mediterranean area: Contrasting Lagrangian and Eulerian models for moisture sources identification. Hydrol. Earth Syst. Sci., 25, 64656477, https://doi.org/10.5194/hess-25-6465-2021.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, https://doi.org/10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, L. Gimeno, and T. Ambrizzi, 2008: A Lagrangian identification of major sources of moisture over central Brazil and La Plata Basin. J. Geophys. Res., 113, D14128, https://doi.org/10.1029/2007JD009547.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, and L. Gimeno, 2011a: On the contribution of the Tropical Western Hemisphere Warm Pool source of moisture to the Northern Hemisphere precipitation through a Lagrangian approach. J. Geophys. Res., 116, D00Q04, https://doi.org/10.1029/2010JD015397.

    • Search Google Scholar
    • Export Citation
  • Drumond, A., R. Nieto, E. Hernandez, and L. Gimeno, 2011b: A Lagrangian analysis of the variation in moisture sources related to drier and wetter conditions in regions around the Mediterranean Basin. Nat. Hazards Earth Syst. Sci., 11, 23072320, https://doi.org/10.5194/nhess-11-2307-2011.

    • Search Google Scholar
    • Export Citation
  • Durán-Quesada, A. M., L. Gimeno, J. A. Amador, and R. Nieto, 2010: Moisture sources for Central America: Identification of moisture sources using a Lagrangian analysis technique. J. Geophys. Res., 115, D05103, https://doi.org/10.1029/2009JD012455.

    • Search Google Scholar
    • Export Citation
  • Durán-Quesada, A. M., L. Gimeno, and J. Amador, 2017: Role of moisture transport for Central American precipitation. Earth Syst. Dyn., 8, 147161, https://doi.org/10.5194/esd-8-147-2017.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., and M. Živković-Rothman, 1999: Development and evaluation of a convection scheme for use in climate models. J. Atmos. Sci., 56, 17661782, https://doi.org/10.1175/1520-0469(1999)056%3C1766:DAEOAC%3E2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., R. Nieto, R. Trigo, S. M. Vicente-Serrano, and J. I. López-Moreno, 2010a: Where does the Iberian Peninsula moisture come from? An answer based on a Lagrangian approach. J. Hydrometeor., 11, 421436, https://doi.org/10.1175/2009JHM1182.1.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., A. Drumond, R. Nieto, R. M. Trigo, and A. Stohl, 2010b: On the origin of continental precipitation. Geophys. Res. Lett., 37, L13804, https://doi.org/10.1029/2010GL043712.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., and Coauthors, 2012: Oceanic and terrestrial sources of continental precipitation. Rev. Geophys., 50, RG4003, https://doi.org/10.1029/2012RG000389.

    • Search Google Scholar
    • Export Citation
  • Gimeno, L., and Coauthors, 2020: Recent progress on the sources of continental precipitation as revealed by moisture transport analysis. Earth-Sci. Rev., 201, 103070, https://doi.org/10.1016/j.earscirev.2019.103070.

    • Search Google Scholar
    • Export Citation
  • Gómez-Hernández, M., A. Drumond, L. Gimeno, and R. Garcia-Herrera, 2013: Variability of moisture sources in the Mediterranean region during the period 1980-2000. Water Resour. Res., 49, 67816794, https://doi.org/10.1002/wrcr.20538.

    • Search Google Scholar
    • Export Citation
  • Grythe, H., N. I. Kristiansen, C. D. Groot-Zwaaftink, S. Eckhardt, P. Tunved, R. Krejci, J. Ström, and A. Stohl, 2017: A new aerosol wet removal scheme for the Lagrangian particle model FLEXPART v10. Geosci. Model Dev., 10, 14471466, https://doi.org/10.5194/gmd-10-1447-2017.

    • Search Google Scholar
    • Export Citation
  • Hanna, S. R., 1982: Applications in air pollution modeling. Atmospheric Turbulence and Air Pollution Modelling, F. T. M. Nieuwstadt and H. Van Dop, Eds., D. Reidel Publishing Company, 275–310.

  • Hersbach, H., and D. Dee, 2016: ERA5 reanalysis is in production. ECMWF Newsletter, No. 147, ECMWF, Reading, United Kingdom, 7, http://www.ecmwf.int/sites/default/files/elibrary/2016/16299-newsletter-no147-spring-2016.pdf.

  • Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 19992049, https://doi.org/10.1002/qj.3803.

    • Search Google Scholar
    • Export Citation
  • Hong, S. Y., and J. O. J. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6). J. Korean Meteor. Soc., 42, 129151.

  • Hong, S. Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 23182341, https://doi.org/10.1175/MWR3199.1.

    • Search Google Scholar
    • Export Citation
  • Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long–lived greenhouse gases: Calculations with the AER radiative transfer models. J. Geophys. Res., 113, D13103, https://doi.org/10.1029/2008JD009944.

    • Search Google Scholar
    • Export Citation
  • Insua-Costa, D., and G. Miguez-Macho, 2018: A new moisture tagging capability in the Weather Research and Forecasting model: Formulation, validation and application to the 2014 Great Lake-effect snowstorm. Earth Syst. Dyn., 9, 167185, https://doi.org/10.5194/esd-9-167-2018.

    • Search Google Scholar
    • Export Citation
  • Insua-Costa, D., G. Miguez-Macho, and M. C. Llasat, 2019: Local and remote moisture sources for extreme precipitation: A study of the two catastrophic 1982 western Mediterranean episodes. Nat. Hazards Earth Syst. Sci., 23, 38853900, https://doi.org/10.5194/hess-23-3885-2019.

    • Search Google Scholar
    • Export Citation
  • Jimenez, P. A., J. Dudhia, J. F. Gonzalez-Rouco, J. Navarro, J. P. Montavez, and E. Garcia–Bustamante, 2012: A revised scheme for the WRF surface layer formulation. Mon. Wea. Rev., 140, 898918, https://doi.org/10.1175/MWR-D-11-00056.1.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170181, https://doi.org/10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NMC/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437471, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lavers, D. A., and G. Villarini, 2013: The nexus between atmospheric rivers and extreme precipitation across Europe. Geophys. Res. Lett., 40, 32593264, https://doi.org/10.1002/grl.50636.

    • Search Google Scholar
    • Export Citation
  • Lavers, D. A., G. Villarini, R. P. Allan, E. F. Wood, and A. J. Wade, 2012: The detection of atmospheric rivers in atmospheric reanalyses and their links to British winter floods and the large-scale climatic circulation. J. Geophys. Res., 117, D20106, https://doi.org/10.1029/2012JD018027.

    • Search Google Scholar
    • Export Citation
  • Lehner, B., K. Verdin, and A. Jarvis, 2008: New global hydrography derived from spaceborne elevation data. Eos, Trans. Amer. Geophys. Union, 89, 9394, https://doi.org/10.1029/2008EO100001.

    • Search Google Scholar
    • Export Citation
  • Lionello, P., and Coauthors, 2006: The Mediterranean climate: An overview of the main characteristics and issues. Dev. Earth Environ. Sci., 4, 126, https://doi.org/10.1016/S1571-9197(06)80003-0.

    • Search Google Scholar
    • Export Citation
  • Miguez-Macho, G., G. L. Stenchikov, and A. Robock, 2004: Spectral nudging to eliminate the effects of domain position and geometry in regional climate model simulations. J. Geophys. Res., 109, D13104, https://doi.org/10.1029/2003JD004495.

    • Search Google Scholar
    • Export Citation
  • Nieto, R., L. Gimeno, and R. M. Trigo, 2006: A Lagrangian identification of major sources of Sahel moisture. Geophys. Res. Lett., 33, L18707, https://doi.org/10.1029/2006GL027232.

    • Search Google Scholar
    • Export Citation
  • Nieto, R., L. Gimeno, A. Drumond, and E. Hernandez, 2010: A Lagrangian identification of the main moisture sources and sinks affecting the Mediterranean area. WSEAS Trans. Environ. Dev., 6, 365374.

    • Search Google Scholar
    • Export Citation
  • Nogueira, M., 2020: Inter-comparison of ERA-5, ERA-Interim and GPCP rainfall over the last 40 years: Process-based analysis of systematic and random differences. J. Hydrol., 583, 124632, https://doi.org/10.1016/j.jhydrol.2020.124632.

    • Search Google Scholar
    • Export Citation
  • Numaguti, A., 1999: Origin and recycling processes of precipitating water over the Eurasian continent: Experiments using an atmospheric general circulation model. J. Geophys. Res., 104, 19571972, https://doi.org/10.1029/1998JD200026.

    • Search Google Scholar
    • Export Citation
  • Peixoto, J. P., M. De Almeida, R. D. Rosen, and D. A. Salstein, 1982: Atmospheric moisture transport and the water balance of the Mediterranean Sea. Water Resour. Res., 18, 8390, https://doi.org/10.1029/WR018i001p00083.

    • Search Google Scholar
    • Export Citation
  • Pisso, I., and Coauthors, 2019: The Lagrangian particle dispersion model FLEXPART version 10.4. Geosci. Model Dev., 12, 49554997, https://doi.org/10.5194/gmd-12-4955-2019.