Assessing the influence of El Niño on the California precipitation regime during the satellite precipitation era

A strong El Niño condition is likely to bring heavy precipitation over California. However, the 2015/2016 El‐Niño generally considered similar and comparable to another extreme event—the 1997/1998 El Niño—failed to bring above‐average precipitation over California. The muted influence of the 2015–2016 El‐Niño on California rainfall has renewed interest in the relationship between El Niño and California precipitation variability. This study assesses seasonal and regional aspects of the precipitation regimes in California for the past two decades using primarily satellite‐based precipitation products. Two satellite‐based precipitation datasets—Integrated Multi‐satellitE Retrievals for GPM (IMERG) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks‐Climate Data Record (PERSIANN‐CDR)—along with the ground‐based measurement—Global Precipitation Climatology Centre (GPCC) dataset as a reference—are analysed to understand the seasonal and regional influence of El‐Niño on precipitation regimes during the 2001–2019 period using Multivariate El Niño Southern Oscillation Index (MEI). The results show that El Niño's impact on precipitation rate varies in different seasons and regions. Specifically, we observe a statistically significant strong negative correlation (−0.72 to −0.66) between MEI and precipitation rate over southern California during the El Niño months. Both satellite precipitation products and ground‐based precipitation measurements indicate a similar El Niño–precipitation relationship over California during the study period. These results contradict the pre 2000 ‘classic’ story that El Niño was positively correlated with higher California precipitation.

important to understand the atmospheric and oceanic conditions that may alleviate or end the prolonged running droughts.
During El Niño, the warm phase of the El Niño Southern Oscillation (ENSO) cycle, trade winds are weakened, and warm water is pushed back east, towards the west coast of the Americas.The warmer waters cause the Pacific jet stream to move south of its neutral position.With this shift, areas in the northern United States and Canada experienced dryer and warmer climate than usual.But in California, the Gulf Coast and the southeast, these periods have been wetter than expected and have experienced increased flooding (NOAA, 2021).During the El Niño events of 1998 and earlier, California received above normal precipitation (Andrews et al., 2004;DeFlorio et al., 2013;Hoerling & Kumar, 1997;Kao & Yu, 2009;Schonher & Nichilson, 1989;Yoon et al., 2015).Although only about 7%-8% of the variation in precipitation anomalies is associated with El Niño, it is capable of alleviating or ending long-lasting droughts in California (Savtchenko et al., 2015).The relationship between El Niño and above-average precipitation over California has been varying from north to south; it depended significantly on the strength of El Niño (Hoell et al., 2016;Jong et al., 2016).Out of eight major El Niño events between 1950 and 2019, four events resulted in a median or above-median precipitation over the entire state.The rest provided a small surplus in precipitation over Southern parts of the state (NOAA, 2015).Many investigations have explored El Nino's relation to anomalous precipitation over California.Schonher and Nichilson (1989) concluded that annual precipitation was above average in California during El Niño episodes between 1951 and 1978 and during the winter of 1982/1983 (Schonher & Nichilson, 1989).Redmond and Koch (1991) investigated the correlation between October/March precipitation and the mean Southern Oscillation Index (SOI) for the preceding June to November and concluded that coastal California received increased precipitation during El Niño like conditions and decreased precipitation during La Niña conditions during 1931-1984 period.Cayan et al. (1999) shows that the frequency of 90th percentile precipitation and streamflow was greatly enhanced across coastal southern California (SCA) during the El Niño phase versus the La Niña phase, and the relative precipitation enhancement diminished northward and vanished along with coastal Northern California (NCA) during 1948-1995 period (Cayan et al., 1999).Hoell et al. (2016) examined California's precipitation sensitivity to El Niño of various intensities and concluded that probability of getting statewide wetter conditions was higher during strong El Niño compared with weak to moderate events for the period of 1896-2014.
What happened in 2015/2016?The strong positive sea surface temperature anomalies (SSTA) started to emerge by the summer of 2015, in March 2015, NOAA forecasters declared the official El Niño (NOAA, 2015).These El Niño conditions were expected to end longrunning drought conditions in the state that started in 2011.Prior to 2015/2016, strong El Niño events resulted in wetter conditions over California. This included the 1982/1983and 1997/1998 El Niño events that brought heavy flooding over the entire state (Hoell et al., 2016).However, 2015/2016 El Niño, one of the strongest events ever recorded compared to other extreme events, failed to bring wetter conditions over California (Cash & Burls, 2019;Lee et al., 2018;Siler et al., 2017;Zhang et al., 2018).The muted influence of 2015/2016 El Niño renewed our interest in studying the effect of El Niño on the precipitation regime of California during the 21st century-an era of satellite precipitation products.
In this study, we focus on reassessing El Niño's influence on the California precipitation regime based on historical precipitation observations and the ENSO index.In contrast to previous studies, we have used satellite precipitation products and a ground-based precipitation product to assess the relationship between the California precipitation and multivariate ENSO index (MEIv2) and shed light on the capability of satellite precipitation products to provide accurate precipitation for the El Niño studies.The objectives of this study is to, (1) examine the El Niño-California precipitation relationship based on seasons, regions and the intensity of the El Niño from 2001 to 2019 and (2) assess the El Niño-precipitation relationship using the satellite-based precipitation products in comparison to a groundbased precipitation product.To our knowledge, this is the first study to assess the effects of El Niño on the precipitation regime using satellite-based precipitation products.The finding of this study fundamentally challenges the longstanding pre-2000 year paradigm, which asserted a strong positive correlation between El Niño occurrences and heightened precipitation levels in California (Cash & Burls, 2019;Cayan et al., 1999;Hoell et al., 2016;Lee et al., 2018;Schonher & Nichilson, 1989;Siler et al., 2017;Zhang et al., 2018).

| DATA AND METHODS
This study assessed the relationships between California precipitation and El Niño during the 2001-2019 period using precipitation datasets and the ENSO index.The three precipitation datasets are taken from the Global Precipitation Climatology Centre (GPCC), the Integrated Multi-satellitE Retrievals for GPM (IMERG), and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR).As shown in Figure 1, California is divided into northern and southern parts based on the characteristics of the annual climatological precipitation.NCA is defined as the region within 36 N-42 N, while SCA is within 32 N-36 N.
GPCC's Full Data Monthly precipitation dataset is a groundbased product based on anomalies from the climatological normal at the stations.GPCC database contains data from more than 124 000 different stations, its application is recommended for hydrometeorological model verification and water cycle studies (Schneider et al., 2022).Precipitation data for GPCC is available at monthly scale and it is internally interpolated on a 0.25 spatial resolution.The Full Data Monthly product of the GPCC precipitation dataset is downloaded from the Deutscher Wetterdienst (DWD) website: https://opendata.dwd.de/climate_environment/GPCC/html/download_gate.html(Schneider et al., 2020).The precipitation data are downloaded for the period of 2001-2019 and the precipi- The final run version of the IMERG precipitation product is downloaded from the GES DISC portal: https://disc.gsfc.nasa.gov/,at a spatial resolution of 0.1 and temporal resolution of 1 month (Huffman et al., 2019).The IMERG algorithm is intended to intercalibrate, merge and interpolate satellite microwave precipitation estimates, together with microwave-calibrated infrared satellite estimates, precipitation gauge analyses, and potentially other precipitation estimators at fine time and space scale for the TRMM and GPM eras over the entire globe.The IMERG product is a combination of satellite-based and ground-based dataset, which is validated using GPCC precipitation gauge product (Huffman et al., 2019).We download the IMERG data To understand the seasonal and regional influence of El Niño on the California precipitation regime, we evaluated the relationship between precipitation rate and MEIv2 during early winter, late winter, winter and all months.Winter months are defined from November to April (NDJFMA), early and late winter months are defined as November to January (NDJ) and February to April (FMA), respectively.
To examine the relationship between precipitation and MEIv2 without the influence of underlying trend, we detrended monthly precipitation data by removing linear trend at every grid cell for the data that were available from GPCC, IMERG and PERSIANN-CDR to evaluate the portion of California precipitation variability that is driven by ENSO.
Thus, the monthly detrended data built at every grid cell are regressed with the ENSO index.The Spearman's Rho correlation coefficient between detrended monthly precipitation and MEIv2 ENSO index at every grid cell is computed.Spearman's Rho correlation coefficient where X i is the ranking of precipitation, Y j is the ranking of MEIv2, X is the mean of precipitation ranking and Y is the mean of MEIv2 ranking.SCA.For a precipitation variable 'X' with mean of 'X' and a standard deviation of 'S', the z-score (Z) of a value 'X' is calculated using Equation 2.

| Correlation analysis: detrended monthly precipitation and MEIv2
The monthly average precipitation rate over the entire state, NCA and SCA, is shown in Figure 2. It is clear from the figure that NCA receives up to double the precipitation rate of SCA across every month except July, August, and September.There is a clear peak in total precipitation rate in the winter months, from December to February.Figure 3 shows a spatially varying time correlation of detrended monthly pre-

| Correlation analysis during El Niño months
To further understand the influence of El Niño on California precipitation rate, we calculated the spatially varying time correlation between detrended monthly precipitation and MEIv2 during El Niño months.
As explained in the section 2, El Niño months are identified when MEIv2 exceeds 0.5.We assessed this spatially varying time correlation during all months, NDJFMA, NDJ and FMA months when MEIv2 is above 0.5.As shown in Figure 4

| Correlation during late winter (FMA) months
We calculated the areal average of grid-based precipitation data over California and calculated precipitation z-scores for each dataset as explained in section 2.   (e.g., 1982/1983, 1997/1998) that brought exceptionally wet winter conditions over SCA (Zhang et al., 2018)  Rather, they should serve as a reminder that El Niño forcing may bring different impacts on the SCA winter precipitation.
We observed the differences in precipitation estimates among the three precipitation datasets (Figure 1), which could be a result of algorithm adopted for estimating precipitation.The PERSIANN-CDR estimates are based on artificial neural networks, which are adjusted using GPCP monthly product version 2.2 (Nguyen et al., 2019).The IMERG estimates are based on combined information of GPM satellite constellation, which are adjusted using GPCC (Huffman et al., 2019), this could be a reason behind similar mean precipitation estimates of GPCC and IMERG.Both IMERG and PERSIANN-CDR reply on the precipitation estimation from the satellite data.It should be noted that these satellite precipitation products may be subject to errors and biases, and the methods used to convert satellite observations into precipitation estimates can introduce additional uncertainties.This may include difficulties in accurately capturing certain precipitation types (e.g., light rain or frozen precipitation), which has been reported in a number of literatures (e.g., Hamza et al., 2020;Peng et al., 2021;Pradhan et al., 2022;Tang et al., 2020).Also, the spatial resolution of satellite precipitation products is acknowledged to be limited for local studies (Abdollahipour et al., 2022;Gebremichael et al., 2008;Shen & Yong, 2021).We used the GPCC Full Data Monthly dataset as the ref- erence in this study.It is important to mention that the performance of the GPCC may be affected in this region with complex topography due to the potential influence of topographical gradient on the interpolated surfaces (Schneider et al., 2022;Zandler et al., 2019).This Considering the complex nature of El Niño, characterized by irregular intervals and varying intensities, average oscillation period of ENSO cycle typically varies from 3 to 7 years (Boyne, 2024).The availability of the precipitation datasets used in this study overlaps only after June 2000.The robustness of our correlation analysis between El Niño events and precipitation will be strengthened as satellite record lengths increase and more El Niño events become available.It will be crucial to understand the relationship between future El Niño events and California precipitation.The findings of this research can be used as a basis for future studies to evaluate additional events based on the relationship between El Niño and California precipitation determined in this study.
covering the study period of 2001-2019 with the precipitation unit of 'mm/h'.PERSIANN-CDR provides monthly rainfall estimates at 0.25 for the latitude band 60 N--60 S over the period of 01/01/1983 to present.The PERSIANN-CDR precipitation product is a satellite-based product adjusted using Global Precipitation Climatology Project (GPCP) monthly product throughout the entire record.The monthly PERSIANN-CDR precipitation dataset is downloaded from the Center for Hydrometeorology and Remote Sensing data portal, at a spatial resolution of 0.25 by 0.25(Nguyen et al., 2019).This precipitation dataset is downloaded for the period of 2001-2019.The unit of the precipitation dataset is mm/h.The monthly MEIv2 is gathered from the NOAA PSL website: https://psl.noaa.gov/enso/mei/, to identify El Niño events during the study period.MEIv2 is a dimensionless index based on five variables: sea level pressure, SSTs, 10-m zonal wind, 10-m meridional wind and outgoing longwave radiation(Wolter & Timlin, 1998;Zhang et al., 2019).MEIv2 classifies ENSO cycle into three phases, (1) La Niña (MEIv2 ≤ À0.5), (2) Neutral (À0.5 < MEIv2 < 0.5) and 3) El Niño (MEIv2 ≥ 0.5).In this study, we focused on El Niño phase of ENSO cycle.El Niño months are identified when MEIv2 exceeds 0.5 for that month.Out of 228 months, 43 months are identified as the El Niño months.In this study, the definition of weak, moderate and strong El Niño classification is based on the magnitude of the MEIv2.Weak El Niños are defined for the MEIv2 between 0.5 and 1, moderate El Niños are defined for the MEIv2 between 1 and 1.5, and strong El Niños are defined when the MEIv2 exceeds 1.5 for the corresponding month(Hoell et al., 2016).Based on MEIv2, 28 months are classified as weak El Nino months, 6 months are classified as moderate El Nino months and 9 months are classified as strong El Nino months.
The significance tests for the correlation coefficients are conducted by calculating p-values using permutation distributions for the twotailed test in MATLAB.The significance level is 5% that correlations are considered significant whose p-values are less than 0.05.Precipitation z-scores are calculated using areal average of precipitation rate over CA, NCA and SCA during FMA months to assess effects of El Niño on precipitation using GPCC, IMERG and PERSIANN-CDR datasets.The areal average of precipitation rate is the arithmetic averaging of precipitation rates at grid cells covering the entire state, NCA and F I G U R E 1 Mean precipitation rate over California (a) GPCC, (b) IMERG and (c) PERSIANN-CDR datasets during 01/2001-12/2019.Dash line divides CA to NCA and SCA at 36 N latitude.
cipitation and MEIv2 at each grid point from 2001 to 2019.Overall, the correlation between detrended monthly precipitation and MEIv2 is weak, ranging between À0.11 to 0.38.The weak positive correlation is evident in NCA during the NDJ months (Figure 3g-i).In comparison, SCA is positively correlated during FMA months.The eastern part of SCA is significantly correlated at a 95% confidence level during the FMA months (hatched grid points in Figure 3j-l).
, the NDJ months indicated a moderate to strong positive correlation, while FMA months indicate a moderate to strong negative correlation.Almost the entire SCA shows a statistically significant strong negative correlation during the FMA months (Figure 4j-l).It indicates low precipitation rate during El Niño months over SCA.Although some parts of NCA show moderate to strong positive correlation during NDJ months, it is statistically not significant.This relationship of low precipitation rate during El Nino months holds for the year of 2001 and later.Since the relationship between precipitation rate and MEIv2 is only significant during the FMA El Niño months, we highlight the results of FMA months in the following section.
Figure 5 shows a scatter plot for the precipitation z-scores of CA, NCA and SCA as a function of MEIv2 during the FMA months.The colours in Figure 5 indicate El Niño (red), La Niña (blue) and neutral (white) months according to the NOAA definition.The dotted red line in Figure 5 is a linear fit for El Niño months.During FMA, 11 months were identified as the El Niño months, out of which five El Niño months are classified as the moderate to strong El-Niño.As evident from the figure, there is a negative trend between FMA precipitation z-scores and MEIv2 during El Niño for the entire state.Particularly, SCA's trend is statistically significant and shows a strong negative correlation between precipitation z-scores and MEIv2 during FMA.Results indicate a negative trend between precipitation rate and El Niño events over entire California during FMA.This relationship is stronger and statistically significant over SCA (r ranges from À0.72 to À0.66) and weakens northward.

3. 4 |
Comparing precipitation-El Niño relationship among GPCC, IMERG and PERSIANN-CDR The mean precipitation rate during 2001-2019 over California indicates minor differences across estimates of GPCC, IMERG and PERSIANN-CDR precipitation products.Compared with the ground-based precipitation measurements of GPCC, PERSIANN-CDR estimates indicate diminished precipitation rates over the northwest of California, while the IMERG estimates are nearly similar to the GPCC estimates (Figure 1).Correlation analysis of precipitation rate and MEIv2 shows similar seasonal and regional trends for GPCC, IMERG and PERSIANN-CDR (Figures 3 and 4).Particularly during FMA, GPCC, IMERG and PERSIANN-CDR indicate significant correlations over SCA during El Niño months (Figure 4j-l).F I G U R E 2 Monthly average precipitation rate over CA, NCA and SCA during 01/2001-12/2019.NCA, northern California; SCA, southern California.F I G U R E 3 Correlation between detrended monthly precipitation and MEIv2 during all months (a-c), NDJFMA (d-f), NDJ (g-i) and FMA months (j-l) for GPCC (a, d, g and j), IMERG (b, e, h and k) and PERSIANN-CDR (c, f, i and l) for the time period of 2001-2019.Hatched grid points indicate significant correlation at the significance level of 5%.F I G U R E 4 Correlation between detrended monthly precipitation and MEIv2 for only El Niño months (a-c), NDJFMA (d-f), NDJ (g-i) and FMA months (j-l) for GPCC (a, d, g and j), IMERG (b, e, h and k) and PERSIANN-CDR (c, f, i and l) for the time period of 2001-2019.Hatched grid points indicate significant correlation at the significance level of 5%.F I G U R E 5 Precipitation z-scores of CA, NCA and SCA as a function of MEIv2 during 01/2001-12/2019 during FMA months.(a) CA GPCC, (b) NCA GPCC, (c) SCA GPCC, (d) CA IMERG, (e) NCA IMERG, (f) SCA IMERG, (g) CA PERSIANN-CDR, (h) NCA PERSIANN-CDR and (i) NCA PERSIANN-CDR.Red, blue and white markers represent El Niño (n = 11), La Niña (n = 21) and neutral (n = 25) months, respectively.Red dash line is a linear fit for the El Niño months.Statistically significant fit is marked with '*'.NCA, northern California; SCA, southern California.F I G U R E 6 Time-series of precipitation z-scores of (a) GPCC, (b) IMERG and (c) PERSIANN-CDR over southern California, and (d) MEIv2 during the 01/2001-12/2019, where red, blue and white bars represent El Niño, La Niña and neutral months, respectively.Pink shaded area indicates major El Niño events during the study period.4| DISCUSSION AND CONCLUSION In this study, we assessed El Niño-precipitation relationship over California using satellite precipitation measurements and ENSO index from 2001 to 2019.We identified and classified four El Niño event occurrences based on their intensity during our study period (highlighted in Figure 6).Based on MEIv2 values, the El Niño event of 2002/2003 and 2006/2007 is classified as weak events, the 2009/10 event is classified as a moderate event and the 2015/16 El Niño event, which persisted for consecutive 13 months with an average MEIv2 value of 1.7, is classified as a strong event.In fact, previous studies have identified the 2015/2016 El Niño event as a 'Super El Niño' comparable with historic strong El Niño events story of El Niño strongly controlling California precipitation breaks down when only the period after 2001 is considered.In order to strengthen confidence in the CA precipitation-El Niño relationship, a deeper understanding of the physical mechanisms by which El Niño affects California rainfall is required.The location and pattern of El Niño in the equatorial pacific play an important role in determining its influence on California's precipitation regime.The major difference between the two most recent strong El Niño events, 1997/98 and 2015/16, can be attributed to the location and distribution of warmer SST in the equatorial Pacific Ocean(Paek et al., 2017).During the 1997/1998 El Niño event, positive SST anomalies were present in the equatorial eastern Pacific Ocean, adjacent to the western coast of the Americas.In contrast, during the 2015/2016 El Niño event, positive SST anomalies occurred in the equatorial central Pacific Ocean, away from the western coast of the Americas (Paek et al., 2017).Lee et al. (2018) classified the 1982/1983 and 1997/1998 events as persistent and the 2015/2016 El Niño events as a mix of transitioning-persistent El Niño based on the pattern of occurrence in the Pacific Ocean and emphasized that only persistent El Niño events result in wetter conditions over SCA.This hypothesis dovetails with our analysis of below-average precipitation rate from the 2015/2016 El Niño events and helps explain the results of negative trendline of precipitation during El Niño months over SCA shown in Figure 5. Since each El Niño event is unique, neither the extreme wetness in 1998 nor the extreme dryness in 2016 should be used as a single measure of what to anticipate from a strong El Niño event.

T A B L E 1
Abbreviation: SCA, southern California.a Precipitation Z scores for SCA.
study uses two satellite-based precipitation datasets-IMERG and PERSIANN-CDR, in comparison to the ground-based precipitation dataset-GPCC, to assess correlation between El Niño and precipitation rate over CA during 2001-2019.Our results suggest that these three precipitation datasets show similar trends and correlation between precipitation rate and MEIv2, which shows that IMERG and PERSIANN-CDR could be reliable datasets for assessment of the precipitation-ENSO relationship.One of the limitations of this study is the relatively short time span of the study.
(Cash & Burls, 2019;Jong et al., 2016;Lee et al., 2018;Zhang et al., 2019)ed the similar wet conditions over SCA from the 2015/16 El Niño event as a potential to end a long-running drought that started in 2011(Cash & Burls, 2019;Jong et al., 2016;Lee et al., 2018;Zhang et al., 2019).Our results indicate that except for the 2002/2003 El Niño event,California received below-average precipitation rate during all other ElNiño events(e.g., 2006/2007, 2009/2010 and 2015/2016).In fact, only 40% of the El Niño months (MEIv2 >0.5, red bars in Figure6) during the study period resulted in above-average precipitation rate over SCA.Table1summarizes the El Niño months based on intensities and precipitation z-scores, only three of nine strong El Niño