Atmospheric rivers (ARs) play an important role in the total annual precipitation regionally and globally, delivering precious freshwater to many arid/semiarid regions. On the other hand, they may cause intense precipitation and floods with huge socioeconomic effects worldwide. In this study, we investigate AR-related precipitation using 18 years (2001–2018) of globally gridded AR locations derived from Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). AR precipitation features are explored regionally and seasonally using remote sensing (Integrated Multi-satellitE Retrievals for GPM version 6 [IMERG V6], daily Global Precipitation Climatology Project version 1.3 [GPCP V1.3], bias-adjusted CPC Morphing Technique version 1 [CMORPH V1], and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks [PERSIANN-CDR]) and reanalysis (MERRA-2 and ECMWF Reanalysis 5th Generation [ERA5]) precipitation products. The results show that most of the world (except the tropics) experience more intense precipitation from AR-related events compared to non-AR events. Over the oceans (especially the Southern Ocean), the contribution of ARs to the total precipitation and extreme events is larger than over land. However, some coastal areas over land are highly affected by ARs (e.g., the western and eastern United States and Canada, Western Europe, North Africa, and part of the Middle East, East Asia, and eastern South America and part of Australia). Although spatial correlations for pairs of IMERG/CMORPH and GPCP/PERSIANN-CDR are fairly high, considerable discrepancies are shown in their estimation of AR-related events (i.e., overall IMERG and CMORPH show a higher fraction of AR-related precipitation). It was found that the degree of consistency between reanalysis and satellite-based products is highly regionally dependent, partly due to the uneven distribution of in situ measurements.
- Eighteen years of global AR locations are used to regionally and seasonally assess precipitation features from various precipitation products
- Over the Southern Ocean, precipitation products show the largest inconsistency in capturing ARs contribution to total and extreme precipitation
- The highest inconsistencies between satellite-based and reanalysis precipitation products are seen over the Middle East and northern Africa
Atmospheric rivers (ARs) are elongated narrow corridors of water vapor transport in the low-level jet layer of the atmosphere. Zhu and Newell (1994) proposed this term to reflect their narrowness and importance to the global water cycle for the first time. They found that at any given moment, there are about three to five ARs in each hemisphere which comprise more than 90% of the total meridional water vapor flux in less than 10% of the zonal circumference in the global middle to high latitudes (Zhu & Newell, 1998). ARs are generally located ahead of cold fronts in the warm sector of extratropical cyclones (Ralph & Dettinger, 2011). ARs are typically longer than 2,000 km and less than 1,000 km wide and are usually made of poleward and lateral moisture transport (Dacre et al., 2015; Ralph et al., 2006).
ARs are often considered important for their extreme precipitation. The enhancement of their precipitation rate tends to occur when the warm moist content hits the orography (Dacre et al., 2015). Several studies have shown the key contribution of ARs to total and extreme precipitation along the West and East Coasts of the United States and the west coast of Europe, and their close connection to flooding. Lavers and Villarini (2015) highlighted this issue by investigating the fraction of AR-related precipitation (1979–2012) across Europe and the United States. They found that AR precipitation is dependent on the month of the year, and in winter, the ratio of AR precipitation to total precipitation reaches 30–50%. They also mentioned a decrease in AR contribution to total precipitation over the Mediterranean region and the central United States. In their previous study (Lavers & Villarini, 2013), they identified some parts of Europe where 60–80% of the top 10 annual maximum daily precipitation (between 1979 and 2011) are related to ARs. Lamjiri et al. (2017) estimated that AR contribution to the US West Coast is between 30% and 50% of annual precipitation based on hourly observation from 1948 to 2002, consistent with earlier findings by Guan et al. (2010) and Dettinger et al. (2011). They also indicated that 60–100% of extreme storms on the West Coast with return periods longer than 2 years are linked to ARs. Landfalling ARs produced more than 250 mm of rain over coastal mountains in 2004 in only 60 hr (16–18 February), and all seven floods during 1998–2006 in California's Russian River are linked to the presence of ARs and their heavy rainfall (Ralph et al., 2006). About 40% of precipitation greater than 100 mm/day is associated with ARs in the Southeast United States (Mahoney et al., 2016). In these contexts, the global reaches and impacts of ARs on extreme precipitation (and extreme winds) were recently highlighted (Waliser & Guan, 2017).
Satellite products have played key roles in AR studies (e.g., Guan et al., 2010; Matrosov, 2013; Neiman et al., 2008; Ralph et al., 2004). Recently, Behrangi et al. (2016) investigated the precipitation rate and pattern of widely used satellite-based precipitation products over AR events for a decade from 2003 to 2012 in the North American West Coast. The study concluded that there is “underestimation over land compared to ground observation” in these satellite products and that these products are more consistent over the ocean than land. In their study, the inventory of landfalling AR dates on the west coast of North America (Neiman et al., 2008) was derived from satellite-observed integrated water vapor (IWV) using the set of AR criteria developed in (Ralph et al., 2004). Later, Wen et al. (2018) showed that satellite products and ground weather radar have difficulties in quantifying peak precipitation rates during extreme events compared to ground measurements.
Recently, a long-term, 6-hourly database of global AR locations based on reanalysis products has become available, which is produced from the vertically integrated water vapor transport (IVT) and more sophisticated criteria for AR detection (Guan & Waliser, 2015; Guan et al., 2018). Relative to IWV, IVT incorporates information on dynamics via the inclusion of wind, which better aligns with the definition of ARs as enhanced transport of moisture, rather than moisture content itself (AMS Glossary of Meteorology, 2017). This global, long-term database provides an excellent opportunity to expand the West-Coast-focused study of Behrangi et al. (2016) to assess the performance of satellite precipitation products in capturing AR precipitation across global land and oceans. Here, we cross-compare AR precipitation from four widely used global precipitation products: the Integrated Multi-satellitE Retrievals for GPM version 6 (IMERG V6), the daily Global Precipitation Climatology Project version 1.3 (GPCP V1.3), the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN-CDR), and the NOAA CPC Morphing Technique version 1 (CMORPH V1). Furthermore, we compare the results with the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) reanalysis, and the ECMWF Reanalysis 5th Generation (ERA5) reanalysis to explore potential differences between satellite and reanalysis products.
Consequently, it is expected that the result of this study will help shed light on the impact and potential consequences of selecting different precipitation products (from data sets covered in this study) for flood alarming systems, streamflow simulations (Behrangi et al., 2014), and AR-related studies in any given part of the globe.
2 Data Set and Study Area
2.1 AR Database
The AR database used (https://ucla.box.com/ARcatalog) detects ARs based on three sets of requirements for IVT, including intensity, direction, and geometry, which are discussed in detail in Guan and Waliser (2015) with refinements in Guan et al. (2018) to help detect ARs with less regular shapes. The criteria used for AR detection are as follows: (a) at each grid cell, IVT intensity must be greater than max (85th percentile, 100 kg/m/s), whichever is larger; (b) mean IVT over the AR should be within 45° of AR shape orientation and with an appreciable poleward component (i.e., 50 kg/m/s); (c) AR length must be greater than 2,000 km and the ratio of length to width greater than 2; and (d) as the refinement for less well-structured ARs, requirement (a) is repeated for up to five times, each time with an increase of 2.5 in the IVT percentile threshold if requirements (b) and (c) fail. The AR data used are based on the MERRA-2 reanalysis, offering 0.5° × 0.625°, 6-hourly record over 30 years (1980–2019).
2.2 Precipitation Products
Precipitation data sets used in this study are summarized in Table 1 and can be categorized into two groups: remotely sensed and reanalysis products.
- Remotely sensed products include IMERG V06, CMORPH V1, GPCP V1.3, and PERSIANN-CDR. The first two products are designed to utilize the highest quality precipitation observations in their merging process; thus, they mainly utilize precipitation estimates from passive microwave (PMW) sensors and IR data to either fill the PMW gaps (e.g., in IMERG) or to track precipitation location (i.e., in CMORPH V1) for interpolation of PMW along precipitation trajectory. The gaps in IMERG V06 filled by IR are mainly related to grids over snow and ice surfaces (where PMW retrievals are less reliable) or when time distance from the closest PMW overpass is too far, so precipitation estimate from IR outperforms the morphed PMW precipitation estimate. More detailed information about the retrieval process of IMERG and CMORPH can be obtained from Huffman et al. (2019) and Joyce et al. (2004), respectively. GPCP V1.3 and PERSIANN-CDR are mainly focused on producing climate data record that requires special consideration for consistency and continuity of the precipitation products used in their merging process. Therefore, both GPCP V1.3 and PERSIANN-CDR heavily use IR observations for precipitation estimation, as IR from GOES can go back to about four decades. PMW data are also used to calibrate IR precipitation estimates. It should be noted that PERSIANN-CDR is bias adjusted at a monthly scale with GPCP monthly product, so similarities between the products at a monthly scale are expected. More detailed information about GPCP V1.3 and PERSIANN-CDR can be obtained from Adler et al. (2017) and Ashouri et al. (2015), respectively.
- The reanalysis data used here consists of ERA5 and MERRA-2. ERA5 is produced through the Copernicus Climate Change Service (C3S) by combining large amounts of observations into global estimates using 4D-Var data assimilation in CY41R2 of ECMWF's Integrated Forecast System (IFS). ERA5 replaced the ERA-Interim reanalysis which stopped production in August 2019. ERA5 has 2D single-layer data like total precipitation with a native resolution of ~30 km which are used in this study. MERRA-2 is the latest atmospheric reanalysis of the modern satellite era produced by NASA's Global Modeling and Assimilation Office (GMAO). MERRA-2 assimilates additional observations compared to its predecessor, MERRA, and includes updates to the Goddard Earth Observing System (GEOS) model and analysis scheme. The analysis is calculated using a 3DVAR algorithm based on the Gridpoint Statistical Interpolation with a 6-hr update cycle and the first guess at an appropriate time procedure for computing temporally accurate observation-minus-background departures. MERRA-2 has a resolution of 0.5° × 0.625° and 72 hybrid sigma/pressure levels from the surface to 0.01 hPa and uses observation-based precipitation data as forcing for the land surface parameterization. More information about MERRA-2 can be obtained from Bosilovich et al. (2017) and Reichle et al. (2017). C3S (2017) provides more detailed information about ERA5 products.
|Product||Original temporal resolution||Original spatial resolution||Availability going back to||Coverage|
|IMERG V06||30 min||0.1° × 0.1°||2000||90°S–90°Na|
|GPCP daily V1.3||Daily||1° × 1°||1996||90°S–90°N|
|CMORPH V01||Daily/subdaily||0.25° × 0.25°||1998||60°S–60°N|
|PERSIANN-CDR||Daily||0.25° × 0.25°||1983||60°S–60°N|
|MERRA-2||Hourly||0.5° × 0.625°||1980||90°S–90°N|
|ERA5||Hourly||~0.28° × 0.28°||1979||90°S–90°N|
- a IMERG may have missing data in high latitudes over snow and ice surfaces (see Huffman et al., 2019).
Before analysis, all the precipitation data sets were mapped onto a common daily 1° × 1° resolution maps using the nearest neighborhood method.
In order to improve consistency between the data sets used in this study (i.e., precipitation products and AR masks), all the data sets were re-gridded to daily (if they were not already) and 1° resolutions. ERA5 precipitation estimates were added up from an hourly to a daily time scale, and the AR data set from a 6-hourly to a daily time scale. Similar to Huning et al. (2017, 2019), a grid in the daily map is labeled as an AR if at least one of the four 6-hourly masks indicates an AR for that day for that grid. This may lead to a slight overestimation. However, we found out that not all of the ARs last for the whole 24 hr period, so if we include grids that AR falls in them for all 6-hr time steps, we will underestimate ARs. An 18-year study period (2001–2018, except 2001–2017 for MERRA-2) is considered for this study because the first full year of IMERG starts in 2001. For the sake of cross-comparison between the precipitation products, global coverage is limited to 60°S–60°N latitude in most of the figures and analyses (some figures have full global coverage related to GPCP product). Precipitation intensities are studied pixel based by separating AR and non-AR events. Furthermore, the contribution of ARs to the total precipitation is investigated by using the daily AR masks. Finally, ratios of AR-related to total precipitation rate and AR-related extremes were calculated by adding up the corresponding mask on the time scale of interest (annual or seasonal).
In this study, extreme events are defined as daily precipitation rates larger than the 95th percentile of all daily precipitation rates in the entire year. We study the extreme events both globally at each grid cell and zonally averaged for annual and seasonal time scales using remote sensing and reanalysis products. We also focused on a few selected regions over land where ARs can have a larger impact (e.g., due to floods resulting from extreme precipitation). The regions of interest were selected based on the literature and maps of AR frequency and precipitation intensity produced in this study. Finally, by normalizing AR and non-AR precipitation histograms, the performance of different precipitation products over the selected regions were cross-compared.
The results of this study are presented under two main sections: (1) AR frequency and overall features of AR versus non-AR precipitation and (2) AR-related extreme events.
4.1 AR frequency and overall precipitation features
AR occurrence has a large regional variability as can be seen in Figure 1. It could be more frequent than 100 days (note: an AR occurring over a partial day is still counted as one AR day in our definition; see section 5) per year in some regions, while there might be little to no AR occurrences over some regions of the world. Figure 1 maps the AR frequency of occurrence at each 1° × 1° grid calculated of AR locations. Tropical regions and Antarctica show very low AR frequency (close to 0), while ARs occur most frequently over the Southern Ocean.
Besides AR frequency of occurrence, it is valuable to map the geographical distribution of AR-related precipitation amount and fraction of total precipitation (e.g., Guan & Waliser, 2015). These data are shown in Figure 2 using precipitation estimates from four different satellite products. The first two columns of Figure 2 show the average daily intensity (mm/day) of AR-related precipitation and non-AR precipitation. Ratios of the mean precipitation intensity of AR over non-AR events and the total precipitation amount of AR over non-AR events are also shown in the third and fourth columns, respectively. Overall, AR precipitation intensity is larger than non-AR in most places except over the deep tropics as can be seen in the third column of Figure 2. Compared to the other precipitation products, GPCP indicates less intense precipitation over the tropics (Figures 2b and 2f). While ARs occur much less frequently in the tropics, they show very intense precipitation there (Figure 1). In other words, ARs produce more intense but less frequent rainfall over most of the tropics compared to the other latitudes. Figures 2i–2l indicate that in most of the nontropical regions, the ratio of the AR to non-AR precipitation is larger than 1. The map produced from IMERG data (Figure 2i) shows much larger ratios than other products (Figures 2j–2l), especially over oceans. In some regions (e.g., North Pacific Ocean, Indian Ocean, and part of the North Atlantic Ocean), this ratio suggests that on average, AR precipitation intensity can be four times larger than non-AR precipitation rates. Results from GPCP and PERSIANN-CDR data sets are in good agreement (Figures 2j and 2l), mainly because PERSIANN-CDR is bias adjusted by GPCP at monthly and 2.5° × 2.5° resolution (Ashouri et al., 2015).
On the other hand, ratios of the volume of AR to non-AR precipitation show different patterns (Figures 2m–2p) compared to the precipitation intensity ratios (Figures 2i–2l). The data suggest that most regions over land receive a greater fraction of annual precipitation from non-AR events than AR events, except a few coastal regions such as the western and eastern United States and Canada, Western Europe, North Africa, East Asia, Eastern South America, and parts of the Middle East and Australia. Over most of the ocean poleward of ~30° latitude, AR precipitation volume exceeds that of non-AR as indicated by volume ratios greater than 1 (Figures 2m–2p).
Figure 3 shows a geographical map of the fraction of annual AR precipitation volume that contributes to the total precipitation. Similar to Figure 2 (the last column from left), IMERG and CMORPH show considerably larger fractions than other products over oceans. Figure 3 also clearly shows areas over lands where annual precipitation from AR exceeds that from non-AR events. Due to their importance, here we select a key subset of these land areas (shown in Figure 4) for more in-depth analysis.
Table 2 provides a summary of AR/non-AR intensity and the contribution of ARs to total precipitation over selected regions shown in Figure 4. Based on this table, South America and East Asia experience the most intense precipitation of about 15 mm/day (average of four data sets), and the West Coast is in second place with 11 mm/day in AR's presence. The contribution of ARs to total annual precipitation ranges from 47% over Europe to 62% in South America based on the average of the four products.
|AR intensity (mm/day)||Non-AR intensity (mm/day)||AR/total ratio|
4.2 AR-related extreme events
In this section, extreme precipitation events (those exceeding the 95th percentile) in the presence of ARs are investigated. We will (1) study extreme events globally using the four remotely sensed products explored here, (2) focus on selected regions over land where they are important for applications such as predicting flood and landslide resulting from extreme precipitation, and (3) compare observation and reanalysis products in capturing the extreme events.
The left-side column of Figure 5 shows geographical maps of the percentage of AR-related extremes of precipitation estimates from the four remote sensing products. Correspondingly, the right-side column shows percentage differences from the mean. Tropical regions (including land and ocean) do not experience much of their extremes during AR conditions (less than 15%). In the extratropics, most of the extreme precipitation events are AR-related over the oceans, while over land AR-related precipitation extremes are often only dominant near coastal regions. The percentage of AR-related extreme precipitation events is generally larger for IMERG and CMORPH and smaller for GPCP and PERSIANN-CDR. This might partly be related to the higher resolution of IMERG and CMORPH that is preserved with the nearest neighborhood interpolation. The products also show different patterns over lands. GPCP and PERSIANN-CDR (Figures 5b and 5d) show more inland penetration of ARs over the West Coast of North America and southern South America. CMORPH indicates fewer AR-related extremes over the Middle East (Figure 5c), and CMORPH and IMERG indicate more extremes over Australia (Figures 5a and 5c). The deviation from the multiproduct mean over lands is generally larger outside of the tropics and can exceed 20% (e.g., over the western United States and South America; cannot be observed in Figure 5 due to the colorbar range). White areas in Figure 5 indicate either no AR or no extreme events due to AR presence ever occurred in the study period (northern Africa and parts of the Middle East in CMORPH; Figures 5c and 5g). Furthermore, PERSIANN-CDR (Figures 5d and 5h) shows white areas over the subsidence zones of the South Pacific Ocean (west of South America) and South Atlantic Ocean (west of Africa) that are likely related to missing shallow precipitation. This pattern also shows up in CMORPH but to a much smaller extent. This suggests that both PMW and IR miss some shallow precipitation, but IR misses more than PMW that is also consistent with Behrangi et al. (2012).
The seasonality of AR-related precipitation extremes can be investigated using Figure 6 in which percentage of AR-related extreme precipitation is shown. The very bottom row shows the seasonal mean of the AR-related percentages calculated from the four remote sensing products. Correspondingly, the upper four rows show the differences from mean for each product and each season. There are fewer AR-related extreme events over the Northern Hemisphere (NH) and in boreal summer (Figure 6s), especially over the US West Coast, Eurasia, and Northern Africa. The largest percentage of AR-related extremes in the NH occurs in the wintertime (Figure 6q). The fraction of AR-related extremes is also larger during cold months in the Southern Hemisphere (SH) (Figure 6s). Australia experiences a higher percentage of extremes from June to November (Figures 6s and 6t). There is a good seasonal consistency over the tropics between the remote sensing products used in this study. IMERG shows its highest deviation over the NH in DJF. It indicates values lower than average (in the same season) over North America and parts of Europe and West/East Asia, but higher than average over lower parts of the West/East Coast and North Africa (Figure 6a). It shows less deviation over land in other seasons and almost consistently higher percentages over oceans throughout the year. In contrast, GPCP (thus PERSIANN-CDR) shows little agreement with IMERG in most regions (Figures 6m–6p). CMORPH is closest to IMERG, yet there are obvious differences between them, especially during cold months. This is partly related to the versions of PMW estimates used in the two products and the way they handle precipitation estimates over frozen surfaces. This figure shows that the choice of products selected to study AR-related precipitation extremes matters and may lead to under or overestimation of the role of ARs, regionally, and seasonally.
Table 3 provides a summary of the percentages of AR-related extremes over selected regions (Figure 4) for satellite-based and reanalysis products that are used in this study. Based on the average of all the products (six in total, including ERA5 and MERRA-2), AR-related extremes in the West and East Coasts of the United States in DJF are about 70%, exceeding other regions and seasons. East Asia is in second place, close to 70% in JJA and MAM. Variation in the percentage of AR-related extremes is the highest over the Middle East with 64% in MAM and 9% in JJA which is the driest season in that region.
|Extreme DJF||Extreme JJA|
|Extreme MAM||Extreme SON|
Figure 7 shows normalized volume histogram plots for AR and non-AR precipitation events over the regions displayed in Figure 4. Two plots are presented for each region, one for normalized precipitation volume of AR and the other one for non-AR events; each contains all the precipitation products studied here. For each region, these plots are constructed in three steps: (1) precipitation rates from AR and Non-AR events are collected separately; (2) the number of samples in predefined bins (i.e., 2, 4, 8, 16, 32, 64, 128, 256, 512 mm/day; similar to Behrangi et al., 2012) is counted to construct precipitation frequency histogram; and (3) mean value of precipitation rate at each bin is multiplied by the corresponding frequency histogram (sample count) of the bin. This will give a precipitation volume histogram that is then normalized by dividing precipitation volume at each bin by the cumulative precipitation volume across all of the bins, so the integral of the normalized volume histogram over the entire precipitation intensity range (shown in X-axis) is 1. This plot is helpful because the area under each plot at different intensities can be a measure of the products' tendency to estimate light, moderate, or extreme precipitation events and makes it easier to compare different products for AR and non-AR events in each region. These plots suggest the following:
- For all the studied regions, IMERG precipitation rates for non-AR events (as a comparison agent) fall below IMERG for AR events in the intense part of this histogram, confirming the linkage between the AR presence and very intense precipitation rates.
- Another interesting point is the inconsistency between the reanalysis data sets (MERRA-2 and ERA5) and satellite products particularly over the Middle East, East Asia, Africa, and South America. For remaining areas, data sets seem more consistent. This inconsistency probably originates from a fewer number of in situ data being assimilated into these products over these regions and also more difficulties simulating precipitation timing over land than the ocean (e.g., in representing the diurnal cycle of precipitation).
- PERSIANN-CDR tends to have difficulties in capturing high precipitation rates in some regions. This is a problem for both AR and non-AR events. There are also some similarities between PERSIANN-CDR and GPCP in capturing extreme precipitation rates that might be related to the fact that PERSIANN-CDR is calibrated by GPCP at a monthly time scale. On the other hand, IMERG estimation of extreme precipitation exceeds all other products over the most studied regions, which is consistent with Masunaga et al. (2019). CMORPH follows IMERG in terms of capturing very intense precipitation rates, including those that are AR-related. Reanalysis estimates tend to fall within the range of what satellite products suggest, although there is no systematic relationship between their histograms and those from satellite products.
- For non-AR events, MERRA-2 generally estimates more extreme events followed by CMORPH and IMERG. PERSIANN-CDR captures fewer extreme events and estimates more light precipitation events almost in all regions.
The US West Coast is recognized as an area likely to be highly affected by AR events as mentioned before. The plot of AR events for this region (Figure 7a) shows that the IMERG product estimates higher precipitation rates for AR events than other products. Furthermore, non-AR IMERG data estimate higher light precipitation and lower extreme precipitation compared to IMERG for AR events. The Middle East is another region that is potentially highly affected by AR events (e.g., Akbary et al., 2019) but not well investigated so far in that regard. ARs over the Middle East can produce severe floods that can lead to major loss of lives and properties and evacuation of thousands based on a recent case study (Dezfuli, 2019). Figure 7 clearly shows that AR events tend to produce more extreme precipitation rates than non-AR precipitation events.
Figures 8 and 9 summarize the contribution of ARs to precipitation extremes and total precipitation at bins of 1° latitude between 60°S and 60°N, separately for land and ocean and different seasons. Over land, the percentages of AR-related extremes decrease from higher latitudes (60°N and 60°S) to the tropics and reach 0 at and near the equator (Figures 8a–8d). Figure 8 shows that the zonal distribution of AR-related extreme precipitation events is less symmetric over land than ocean. For example, over land, the SH has a higher percentage of extreme events under AR events, and this percentage can be up to 70% near 40°S latitude in all seasons except spring, but in the NH, the percentages do not exceed 60% and are often below 50%. Over both land and ocean, there is more consistency between the products closer to the tropics. The spread among the products is generally larger at higher latitudes where satellite products face challenges in estimating precipitation (Behrangi et al., 2012). Over the ocean, generally, higher percentages are observed than over land, and in the SH in MAM, the percentages can reach up to 82%. Note that in colder seasons (DJF/MAM in NH and JJA/SON in SH) and over land, the spread of the percentage is larger among the studied products. No such clear pattern is observed over the ocean.
Figure 9 is similar to Figure 8, but it shows the average ratio of AR precipitation to the total precipitation. The percentages of AR to total precipitation follow a similar zonal pattern to that observed for percentages of AR-related precipitation extremes (shown in Figure 8). Over land (Figures 9a–d), the percentages range between 0 and 60 in both hemispheres. The highest inconsistency between the products is in the NH and over land and occurs within 40–60°N and 10–30°N latitudes in DJF (Figure 9a). The inconsistencies are also large over land in the southern winter (JJA) between 25°S and 60°S. Figure 9 also suggests that the spread among the percentages of AR-related precipitation from different products is larger over higher latitude ocean than land. Furthermore, and similar to Figure 8, there is more symmetry between NH and SH percentages over the ocean than land. Over the ocean, the fraction of the AR-related precipitation amount is generally larger than over land across most latitudes and may reach up to 70% in vast areas over the Southern Oceans. Over the ocean, CMORPH and IMERG consistently show larger fractions compared to the other products, and reanalysis products (MERRA2 and ERA5) are fairly consistent. This is not necessarily the case over land. Note that the observed noisiness of the plots in Figures 8 and 9 (i.e., over land in SH poleward of 40°S) is likely related to the small sample size as shown in Figure S1.
ARs contribute significantly to annual total precipitation and extreme precipitation events. Remote sensing precipitation products provide an observational foundation to quantify such contributions globally, but satellite products often use different retrieval techniques and assumptions in their estimate. They may also face common challenges in capturing certain types of precipitation events, such as orographically enhanced precipitation events during ARs in the western United States (Behrangi et al., 2015), which may result in a systematic error for accurate quantification of precipitation. Because of the large hydrological and scientific implications of extreme precipitation events, this study provides a comprehensive comparison between commonly used satellite-based global precipitation products (IMERG V6, GPCP V1.3, CMORPH V1, and PERSIANN-CDR) to assess their temporal and spatial consistency concerning the presence/absence of ARs. Additionally, two widely used reanalysis products (ERA5 and MERRA2) are compared with satellite products.
Geographical distribution of AR precipitation and AR-related extremes indicates that oceans, generally, are more affected by ARs than land, but the impacts of ARs over land are larger. Maps of the fraction of AR precipitation to total precipitation and AR-related extremes specify higher values over coastal regions in the subtropics compared to other regions over land. We focused on a few highly affected regions over land for a more detailed investigation. The results show that the US West/East Coast and East Asia experience about 70% of the extreme events in the presence of ARs in DJF and MAM, respectively. On the other hand, South America and Australia receive about 62% of their precipitation from ARs. Our findings show a larger fraction of AR contribution to annual precipitation for the west of Europe and the US West Coast compared to two related studies. Lavers and Villarini (2015) indicate 30–50% for AR contribution in winter, but this study suggests 51–65% (based on four different products) for the West Coast and 45–49% for the west of Europe. Guan and Waliser (2015) mention over 30% of contribution to total precipitation for a number of extratropical areas. However, the use of different data sets and criteria for identifying ARs (e.g., 6-hourly ARs in Guan & Waliser, 2015) as well as the choice of precipitation product, and the period of study could explain part of the differences in this comparison. Additionally, findings suggest that regions that receive a considerable amount of their total precipitation from ARs also experience a higher fraction of their extreme events under AR events. This pattern is clearly shown in the zonally averaged distribution of related fractions.
Based on the findings of this study, there is a better agreement among the products over the tropics than in higher latitudes. The largest inconsistencies occur over the Southern Ocean where IMERG shows the highest percentage of contribution of ARs to total precipitation and extreme events and consequently the highest deviation from other products used in this study. It is shown that, overall, pairs of IMERG/CMORPH and GPCP/PERSIANN-CDR have higher spatial correlations globally, which is expected given the similarities in their retrieval methods. Seasonality is another important factor for the spread of the products. In colder seasons and over land, there is more inconsistency between products, which is not the case over oceans. This can be partly related to the large uncertainties of remote sensing products in estimating light precipitation and snowfall over cold regions, especially over snow- and ice-covered surfaces.
It should be mentioned that reanalysis products deviate significantly from satellite-based products over a few selected regions over land (e.g., North Africa and the Middle East) that might be related to the lack of sufficient in situ observations for regional calibration of satellite products and assimilation of them into reanalysis products.
Given their importance, a thorough understanding of future changes in intensity and pattern of AR precipitation is needed. The changes are typically inferred from model runs that are often verified by observation or reanalysis products. Furthermore, accurate estimation of AR precipitation is important for several hydrologic applications (e.g., flood forecasting). The present study shows that despite the recent improvements in our observational tools (e.g., new sensors with higher capabilities), there remain inconsistencies between satellite and reanalysis products and also among satellite products themselves. Part of the differences among the satellite products can be related to the instruments or algorithms used in precipitation retrieval. For example, products such as IMERG and CMORPH, that are mainly based on PMW, show fewer differences compared to PERSIANN-CDR or GPCP that are more dependent on IR observation. Nonetheless, further effort to advance our observational products over both land and ocean is needed, so the gained knowledge can more effectively be used to inform regional and global climate models and advance hydroclimatic predictions.
The research described in this paper was carried out at the University of Arizona. Financial support for the University of Arizona is made available partly from NASA MEaSUREs (NNH17ZDA001N-MEASURES) and NASA Weather and Atmospheric Dynamics (NNH19ZDA001N-ATDM) awards.
Data Availability Statement
IMERG V06 and MERRA-2 data were obtained from the Goddard Earth Sciences Data and Information Services Center (GES DISC) (https://disc.gsfc.nasa.gov/). GPCP V1.3 product was obtained online (https://www.ncei.noaa.gov/data/global-precipitation-climatology-project-gpcp-daily/access/). CMORPH V1 data are available online (https://www.ncei.noaa.gov/data/cmorph-high-resolution-global-precipitation-estimates/access/daily/0.25deg/). PERSIANN-CDR is downloaded online (https://www.ncei.noaa.gov/data/precipitation-persiann/access/), and ERA5 is accessed online (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=form). Data sharing does not apply to this article as no new data were created or analyzed in this study.
|jgrd56561-sup-0001-2020JD033021-SI.pdfPDF document, 66.6 KB||
Supporting Information S1
Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
- 2017). Global precipitation climatology project (GPCP) climate data record (CDR), version 1.3 (daily). Retrieved from https://www.ncei.noaa.gov/data/global-precipitation-climatology-project-gpcp-daily/access/
- 2019). Spatio-temporal changes of atmospheric rivers in the Middle East and North Africa region. International Journal of Climatology, 39(10), 3976–3986. https://doi.org/10.1002/joc.6052
AMS Glossary of Meteorology (2017). Atmospheric river. https://glossary.ametsoc.org/wiki/Atmospheric_river
- 2015). PERSIANN-CDR: Daily precipitation climate data record from multisatellite observations for hydrological and climate studies. Bulletin of the American Meteorological Society, 96(1), 69–83. https://doi.org/10.1175/BAMS-D-13-00068.1
- 2014). Satellite-based precipitation estimation and its application for streamflow prediction over mountainous Western U.S. basins. Journal of Applied Meteorology and Climatology, 53(12), 2823–2842. https://doi.org/10.1175/JAMC-D-14-0056.1
- 2015). On the quantification of atmospheric rivers precipitation from space: Composite assessments and case studies over the eastern North Pacific Ocean and the Western United States. Journal of Hydrometeorology, 17(1), 369–382. https://doi.org/10.1175/JHM-D-15-0061.1
- 2016). On the quantification of atmospheric rivers precipitation from space: Composite assessments and case studies over the eastern North Pacific Ocean and the Western United States. Journal of Hydrometeorology, 17(1), 369–382. https://doi.org/10.1175/JHM-D-15-0061.1
- 2012). On the quantification of oceanic rainfall using spaceborne sensors. Journal of Geophysical Research, 117, D20105. https://doi.org/10.1029/2012JD017979
- 2017). Atmospheric water balance and variability in the MERRA-2 reanalysis. Journal of Climate, 30(4), 1177–1196. https://doi.org/10.1175/JCLI-D-16-0338.1
- 2015). How do atmospheric rivers form? Bulletin of the American Meteorological Society, 96(8), 1243–1255. https://doi.org/10.1175/BAMS-D-14-00031.1
- 2011). Atmospheric rivers, floods and the water resources of California. Water, 3(2), 445–478. https://doi.org/10.3390/w3020445
- 2019). Rare atmospheric river caused record floods across the Middle East. Bulletin of the American Meteorological Society, 101(4), E394–E400.
- 2010). Extreme snowfall events linked to atmospheric rivers and surface air temperature via satellite measurements. Geophysical Research Letters, 37, L20401. https://doi.org/10.1029/2010GL044696
- 2015). Detection of atmospheric rivers: Evaluation and application of an algorithm for global studies. Journal of Geophysical Research: Atmospheres, 120, 12,514–12,535. https://doi.org/10.1002/2015JD024257
- 2018). An Intercomparison between reanalysis and dropsonde observations of the total water vapor transport in individual atmospheric rivers. Journal of Hydrometeorology, 19(2), 321–337. https://doi.org/10.1175/JHM-D-17-0114.1
- 2019). GPM IMERG final precipitation L3 1 day 0.1 degree × 0.1 degree V06.
- 2019). Sensitivity of seasonal snowfall attribution to atmospheric rivers and their reanalysis-based detection. Geophysical Research Letters, 46, 794–803. https://doi.org/10.1029/2018GL080783
- 2017). Implications of detection methods on characterizing atmospheric river contribution to seasonal snowfall across Sierra Nevada, USA. Geophysical Research Letters, 44, 10,445–10,453. https://doi.org/10.1002/2017GL075201
- 2004). CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. Journal of Hydrometeorology, 5(3), 487–503. https://doi.org/10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2
- 2017). Hourly storm characteristics along the U.S. West Coast: Role of atmospheric rivers in extreme precipitation. Geophysical Research Letters, 44, 7020–7028. https://doi.org/10.1002/2017GL074193
- 2013). The nexus between atmospheric rivers and extreme precipitation across Europe. Geophysical Research Letters, 40, 3259–3264. https://doi.org/10.1002/grl.50636
- 2015). The contribution of atmospheric rivers to precipitation in Europe and the United States. Journal of Hydrology, 522, 382–390. https://doi.org/10.1016/j.jhydrol.2014.12.010
- 2016). Understanding the role of atmospheric rivers in heavy precipitation in the Southeast United States. Monthly Weather Review, 144(4), 1617–1632. https://doi.org/10.1175/MWR-D-15-0279.1
- 2019). Inter-product biases in global precipitation extremes. Environmental Research Letters, 14(12), 125016. https://doi.org/10.1088/1748-9326/ab5da9
- 2013). Characteristics of Landfalling atmospheric Rivers inferred from satellite observations over the eastern North Pacific Ocean. Monthly Weather Review, 141(11), 3757–3768. https://doi.org/10.1175/MWR-D-12-00324.1
- 2008). Meteorological characteristics and overland precipitation impacts of atmospheric rivers affecting the West Coast of North America based on eight years of SSM/I satellite observations. Journal of Hydrometeorology, 9(1), 22–47. https://doi.org/10.1175/2007JHM855.1
- 2011). Storms, floods, and the science of atmospheric rivers. Eos, Transactions American Geophysical Union, 92(32), 265–266. https://doi.org/10.1029/2011EO320001
- 2004). Satellite and CALJET aircraft observations of atmospheric rivers over the eastern North Pacific Ocean during the winter of 1997/98. Monthly Weather Review, 132(7), 1721–1745. https://doi.org/10.1175/1520-0493(2004)132<1721:SACAOO>2.0.CO;2
- 2006). Flooding on California's Russian River: Role of atmospheric rivers. Geophysical Research Letters, 33, L13801. https://doi.org/10.1029/2006GL026689
- 2017). Land surface precipitation in MERRA-2. Journal of Climate, 30(5), 1643–1664. https://doi.org/10.1175/JCLI-D-16-0570.1
- 2017). Extreme winds and precipitation during landfall of atmospheric rivers. Nature Geoscience, 10(3), 179–183. https://doi.org/10.1038/ngeo2894
- 2018). How well were the early 2017 California Atmospheric River precipitation events captured by satellite products and ground-based radars? Quarterly Journal of the Royal Meteorological Society, 144(S1), 344–359. https://doi.org/10.1002/qj.3253
- 1994). Atmospheric rivers and bombs. Geophysical Research Letters, 21(18), 1999–2002. https://doi.org/10.1029/94GL01710
- 1998). A proposed algorithm for moisture fluxes from atmospheric Rivers. Monthly Weather Review, 126(3), 725–735. https://doi.org/10.1175/1520-0493(1998)126<0725:APAFMF>2.0.CO;2