Global Impacts of El Niño on Terrestrial Moisture Recycling

The global impacts of El Niño on precipitation have been long‐recognized, but more understanding of the mechanisms behind this influence is needed. For instance, previous studies have largely overlooked the potential impacts of El Niño on terrestrial moisture recycling (TMR). We perform a 40‐year forward tracking simulation to derive a global climatology of recycled precipitation and use a composite analysis to investigate how El Niño affects TMR. We identify seven regions where the El Niño impact on TMR is most significant and find that, in these regions, changes in precipitation and TMR are directly related: they increase or decrease together. In addition, we find a marked latitudinal contrast between the Southern Hemisphere, where TMR increases during El Niño, and the Northern Hemisphere and the tropics, where it decreases. Our results indicate that the weakening and strengthening of TMR can be behind precipitation changes caused by El Niño globally.

showing that El Niño and La Niña events exert significant global effects on the asymmetry of the origin (oceanic vs. terrestrial) of continental precipitation anomalies. However, they did not focus on TMR. Some studies have discussed the possible influence of ENSO on TMR but either remained conceptual or did not focus on quantifying this influence at the global scale. For instance, Poveda et al. (2006) suggested conceptually that the ENSO alters TMR in Northern South America. In Africa, Sheen et al. (2017) characterized the Sahel region as highly sensitive to both TMR and Sorí et al. (2017) analyzed moisture sources of precipitation in the Congo river basin while highlighting the need to investigate how the ENSO affects them. In Asia, Y. Zhao and Zhou (2021) analyzed the interannual variability of TMR over the Tibetan Plateau and found that the ENSO explains around 60% of this variability with an asymmetric response by phase. Closest to our study is Sorí et al. (2021), who investigated whether changes in atmospheric moisture contributions from oceanic or terrestrial sources control global ENSO-induced precipitation anomalies. However, they did not focus on or show TMR anomalies caused by El Niño nor discuss these effects in light of recent studies (our meta-analysis).
This study quantifies the long-term impact of El Niño on TMR from a composite analysis and a 40-year simulation of water vapor forward tracking with the Water Accounting Model-2layers (WAM-2layers; van der Ent et al., 2014), using an explicit TMR representation. Precipitation response to El Niño shows strong seasonality (e.g., Wang et al., 2020) andinter-event (e.g., Johnson, 2013) features that are out of the present study's scope. We hypothesize that TMR anomalies are related to El Niño-induced changes in precipitation and identify global regions where this relationship exists. A meta-analysis of known ENSO impacts on regional climates substantiates this global picture. In contrast to previous studies (e.g., Wang-Erlandsson et al., 2018;Weng et al., 2018) and available datasets (Link et al., 2020;Tuinenburg et al., 2020), our tracking experiment is more extended in time, which is important for detecting El Niño-induced effects. The rest of the paper is organized as follows: Section 2 describes our simulation and meta-analysis, Section 3 presents the results while discussing the underlying physical mechanisms, and Section 4 concludes.

Moisture Tracking Simulation
The WAM-2layers  is an Eulerian moisture tracking numerical model. It has been widely used to study atmospheric moisture transport processes worldwide (Findell et al., 2019;Guo et al., 2019;Link et al., 2020;Wang-Erlandsson et al., 2018;D. C. Zemp et al., 2017;C. Zhang et al., 2019) and compares well with other tracking methods (van der Ent et al., 2013). We used the Python version available on the GitHub repository (https://github.com/ruudvdent/WAM2layersPython) and modified the post-processing scripts for our analysis. The moisture tracking algorithm is based on the atmospheric water balance . The atmospheric column is divided into upper and lower levels along a set of layers in an atmospheric model, separating the column into high-altitude and lower-altitude winds. This two-layers approach provides a parsimonious representation of the atmospheric heterogeneity between high-and low-level layers (van der Ent et al., 2013). In the model, precipitation results from two sources of moisture: evapotranspiration from the surface and moisture flux convergence associated with wind circulation (Keys et al., 2014). The model resolves precipitation and evaporation within atmospheric layers through time, differing from Lagrangian backtracking techniques in which moisture sources are evaluated based on relative humidity variations (van der Ent et al., 2013). For a detailed description of the WAM-2layers, we refer the reader to van der Ent et al. (2013van der Ent et al. ( , 2014.
Moisture tracking was conducted forward in time, focusing on the fate of evapotranspiration contributing to TMR to calculate the fraction (recycling ratio, ρr) and amount of precipitation with continental origin (recycled precipitation, Pr). Our tracking experiment spans from 1979 to 2018, discarding the first year for model spin-up (e.g., Link et al., 2020;Wang-Erlandsson et al., 2018). The model domain covers a global grid from 79.5°N to 79.5°S latitude. Calculations use data of specific humidity, zonal and meridional wind speeds at 24 pressure levels, surface pressure at 6-hr intervals, and 3-hr accumulated precipitation and evaporation from the ERA-Interim reanalysis (Dee et al., 2011) on a 1.5° grid with 25,680 cells total (nx = 240, ny = 107). This setup is based on previous studies indicating that it is adequate to track moisture spatially, both forward and backward in time and at regional and global scales (Link et

Composite Analysis
We used the definition of ENSO events described by the Oceanic Niño Index (ONI) to distinguish between El Niño (events) and neutral (non-events) ENSO phases. An El Niño event is characterized by five consecutive overlapping 3-month periods with ONI at or above the +0.5 anomaly. Neutral phases occur when ONI is between 0.5 and −0.5 during five consecutive overlapping 3-month periods ( Figure S1 in Supporting Information S1). The selected El Niño events are first averaged and then subtracted from the average neutral phases to form the composite anomalies (e.g., Welhouse et al., 2016). When assessing differences, statistical significance is determined using a two-tailed Student's t-test with a confidence level of 0.95. These differences are between composites of different ENSO phases at each grid cell, so spatial correlation does not affect the test. Comparisons include our estimates of TMR and precipitation from ERA-Interim.

Meta-Analysis
For climate-related fields, we traced previous studies in peer-reviewed journals tagged as Q1 or Q2 in the SCImago Journal Ranking (SJR). This was done through a manual search using multiple terms and platforms: Google Scholar, Web of Science, and Scopus. This procedure led us to 44 studies (see details in Table S1 in Supporting Information S1) focused on El Niño-induced alterations of precipitation (P) and recycled precipitation (Pr) or related variables: temperature (T), evapotranspiration (E), and vegetation activity (Normalized Difference Vegetation Index [NDVI]). We obtained changes in the analyzed variables from these studies, mainly the sign (positive or negative) in the regions of interest.

Global Patterns of TMR
Our simulation reproduces global patterns of TMR described by previous studies such as Keys et al. (2019) and van der Ent et al. (2010) (Figure 1), including the spatial distribution and magnitude of average recycled precipitation ( Figure 2b). This comparison serves as a validation of our simulation. Recycling ratios are high over the tropical Andes, southern South America, the Sahel, South Africa, and eastern Eurasia (Figure 1), suggesting that changes in TMR can have pronounced effects over these regions. Over North America and western Eurasia, recycling ratios are lower but still considerable (Figure 1). Figure 1. The long-term annual mean of terrestrial moisture recycling (TMR) ratio over 1980-2018 estimated from a Water Accounting Model-2layers simulation. This ratio is the fraction of total precipitation with terrestrial rather than oceanic origin. Figure 2 shows the long-term mean of precipitation, recycled precipitation, and the corresponding anomalies caused by El Niño between 1980 and 2018, based on composites comprising 10 El Niño and 15 neutral phases ( Figure S1 in Supporting Information S1). In regions with the most significant changes in precipitation, these changes are directly related to changes in TMR. We refer to a direct mathematical relationship when both variables increase or decrease together ( Figure S2 in Supporting Information S1). For example, precipitation increases due to El Niño (blue areas in South America and Africa, Figure 2c) coincide with precipitation recycling increases (blue regions in Figure 2d). Likewise, precipitation and recycling decrease in northern North America, eastern Asia, and tropical regions of South America, Africa, and Asia. This direct relationship suggests that changes in recycling are physically related to precipitation anomalies caused by El Niño. We will further discuss this below in light of our meta-analysis.

Regional Patterns of El Niño-Induced Anomalies
Figure 3a shows a regional classification of El Niño-induced anomalies on TMR and precipitation according to the significance level and the sign of change. About 41% of continental regions do not experience significant changes in recycled precipitation associated with El Niño. On the other hand, TMR is significantly reduced (red areas) or increased (blue areas) in 48% and 11% of the continental regions, respectively. These changes show a hemispherically asymmetric pattern: while recycling reduces over vast Northern Hemisphere regions, it increases in Southern Hemisphere areas. Nearly 23% of the continental regions (dark red areas) experience significant reductions in precipitation and TMR during El Niño. In contrast, nearly 8% of the continents experience significant increases in both variables (dark blue areas).  Figure 3b presents the contribution (fraction) of changes in recycled precipitation to changes in total precipitation due to El Niño for regions where both changes are statistically significant (dark blue and dark red areas in Figure 3a). These contributions vary from low values in the tropics and Southern Hemisphere to ∼100% in areas of the Northern Hemisphere. In other words, changes in recycled precipitation from continental sources (TMR) can substantially contribute to changes in (total) precipitation during El Niño in extensive regions of the world (Figure 3b). About 25% of the continental regions show alterations in moisture recycling that do not coincide with precipitation changes (light red areas). This discrepancy may be due to compensating processes; for example, reductions in the precipitation fraction from continental sources (TMR) that are compensated by increased precipitation from In the light blue areas, precipitation and recycled precipitation increase, but the precipitation change is not significant. Changes in both variables on dark blue (increase) and dark red (decrease) areas are significant. In the light red areas, precipitation and recycled precipitation decrease, but the precipitation reduction is not significant. In the gray areas, precipitation recycling is not significantly altered. (b) Percentage ratio between changes in recycled and total precipitation. oceanic sources (e.g., Sorí et al., 2021). These changes might be triggered by moisture convergence increases due to changes in pressure gradients between the ocean and continent (Brönnimann, 2007). Europe offers an example: while previous studies about El Niño effects on precipitation show mixed results (Brönnimann, 2007), our results show reductions in recycled precipitation over the continent's eastern and northern regions (Figure 3). Notably, the extensively cited review by Brönnimann (2007) does not mention recycling when considering the ENSO influence on precipitation in Europe, suggesting that changing TMR is a previously overlooked mechanism behind El Niño's influence in this region.

Linking El Niño Impacts on TMR and Current Knowledge
Based on Figures 2 (see ovals), 3, and 4 highlights seven regions where El Niño relates to significant changes in TMR and precipitation: Northern North America (NNA), NorthEastern Asia (NEA), Northern South America (NSA), the Sahel in Central AFrica (CAF), SouthErn Asia (SEA), Southern South America (SSA), and Southern AFrica (SAF). In these regions, the relationship between precipitation and recycled precipitation changes due to El Niño is direct. Both variables decrease in the tropics and the Northern Hemisphere while increasing in the Southern Hemisphere. Further, previous studies have identified relevant impacts of El Niño in all these regions (Table S1 in Supporting Information S1).
Although our results do not allow establishing a causal relationship between precipitation and recycling changes, the said direct relationship suggests that in these regions, TMR weakening or strengthening contribute, respectively, to precipitation decrease or increase during El Niño. There are physical foundations for speculating this. For instance, El Niño-induced atmospheric circulation alterations can reduce (increase) precipitation, generating terrestrial moisture availability decreases (increases) linked to evapotranspiration decreases (increases). Such evapotranspiration decrease (increase) can weaken (strengthen) TMR, amplifying the initial precipitation change caused by El Niño.
There are four patterns in Figure 4 depending on how El Niño affects the different variables. First, El Niño reduces all variables in NEA (Northern Hemisphere). Second, El Niño reduces all variables except temperature in NNA (also the Northern Hemisphere). Third, El Niño increases all variables in the Southern Hemisphere (SSA and SAF). Fourth, El Niño reduces all variables except temperature and surface moisture divergence in the tropics (NSA, CAF, and SEA). . Schematic representation of El Niño-induced alterations of precipitation (P, hexagon), temperature (T, rhombs), vegetation activity (V, squares), evapotranspiration (E, pentagons), recycled precipitation (Pr, circles), and atmospheric moisture convergence (upward arrows) and divergence (downward arrows). Colors indicate whether a variable decreases (red) or increases (blue). The gray circle indicates a study suggesting that El Niño produces an effect, but the signal is unclear. Numbers identify the studies where the reported results come from (Table S1 in Supporting Information S1). The light colored ovals show El Niño-induced terrestrial moisture recycling and precipitation decreases (red) and increases (blue), as found in this study.

of 10
In NEA (first pattern), recycled precipitation reduction due to El Niño (Figures 2-4) coincides with El Niño-induced moisture convergence and precipitation decreases, likely related to modification of the subtropical jet stream and wave trains along the eastern Asia coast (Lin & Qian, 2019;Wen et al., 2019). In addition, temperature also decreases (Buermann et al., 2003;Lin & Qian, 2019) due to Siberian cold air intrusion related to an anomalous cyclone in the northwestern Pacific during El Niño (J. Li et al., 2017). Further, El Niño affects vegetation dynamics in eastern Russia (J. Li et al., 2017), where temperature and precipitation reductions lead to vegetation activity decreases (L. Zhao et al., 2018) and, therefore, evapotranspiration reductions (J. Li et al., 2017;Martens et al., 2018;Miralles et al., 2014;Yan et al., 2013).
In SSA and SAF (third pattern), recycled precipitation increase due to El Niño (Figures 2-4) coincides with moisture convergence and precipitation increases linked to alterations in atmospheric circulation patterns (Cai et al., 2020;Driver et al., 2019;Hurtado & Agosta, 2020;Pascale et al., 2019). Despite surface radiation reductions linked to precipitation increases, temperature increases due to warmth advection (Cai et al., 2020;Detzer et al., 2020). Although these regions have been identified as energy-limited environments (Martens et al., 2018), observational evidence indicates that vegetation activity and evapotranspiration increase during El Niño (Martens et al., 2018;K. Zhang et al., 2015). Martens et al. (2018) suggest that precipitation increases might still affect the evaporation dynamics in these wet regions through their impact on interception loss.

Conclusions
El Niño has pronounced effects on TMR globally, especially in the regions highlighted in Figure 4. There is spatial variability in these effects, with a marked latitudinal contrast between the Southern Hemisphere, where TMR increases during El Niño, and the Northern Hemisphere and the tropics, where TMR decreases.
El Niño-induced changes in precipitation and TMR have a direct relationship in the most sensitive regions: precipitation and TMR increase or decrease together. However, relationships between TMR changes and other variables (temperature, evapotranspiration, vegetation activity, and moisture convergence or divergence) show differences among regions.
The direct relationship between precipitation and TMR changes during El Niño suggests that TMR weakening or strengthening are mechanisms related to precipitation increases or decreases caused by El Niño. Previous studies about the El Niño impacts on precipitation have largely overlooked these mechanisms. Disentangling the mechanisms through which El Niño affects precipitation via TMR is a future research direction requiring site-specific studies; for instance, to identify whether changes in TMR are driven by changes in vegetation activity, land