Professional Documents
Culture Documents
Nonstationary extreme precipitation in Brazil
Nonstationary extreme precipitation in Brazil
To cite this article: Bianca Nespoli Cortez, Gabrielle Ferreira Pires, Alvaro Avila-Diaz, Humberto
Paiva Fonseca & Lais Rosa Oliveira (2022): Nonstationary extreme precipitation in Brazil,
Hydrological Sciences Journal, DOI: 10.1080/02626667.2022.2075267
DOI: 10.1080/02626667.2022.2075267
Nonstationary extreme precipitation in Brazil
Abstract
Recent studies show that extreme precipitation has changed over the past decades and
increased in intensity and/or frequency on the local, continental and global scales
(Alexander et al., 2006; McPhillips et al., 2018; Westra et al., 2013; Xavier et al., 2019;
Eyring et al., 2021). Although extreme events are associated with natural climate
al., 2016; Lehmann et al., 2015; Porto de Carvalho et al., 2014). In addition, increases in
extreme heavy precipitation are more frequent than decreasing trends (Westra et al.,
2013; Donat et al., 2016; Lehtonen and Jylha 2019), particularly related with the
increase in atmospheric moisture content associated with warming (Eyring et al., 2021) .
behavior and how these changes can impact social and ecological systems. In urban
environments, the impacts may be related to urban mobility, water, energy supply, and
health, especially waterborne diseases. Furthermore, these events are associated with
urban floods and landslides, which are also correlated with accelerated and disorderly
urbanization (McPhillips et al., 2018). Most urban centers in Brazil are the result of a
and geotechnical risk, leaving the population exposed and vulnerable to such impacts.
Densely urbanized areas might be even more vulnerable to extreme events disasters in
the future (Ávila et al., 2016; de Loyola Hummell et al., 2016; Debortoli et al., 2017;
In Brazil, numerous cases of urban floods and landslides are recorded every
year. Recent episodes were recorded in the states of Bahia (in 2021), Minas Gerais and
São Paulo (in 2020 and 2022), Espírito Santo and Rio de Janeiro (2020), Santa Catarina
in 2008, and in some municipalities in Rio de Janeiro (Angra dos Reis in 2009, Niterói
in 2010, and mountaintop cities in 2011) (IBGE, 2019a). All these events resulted in
many deaths and damages to local infrastructure and left thousands of homeless people.
With the intensification of climate change, the population might become even
more exposed, more vulnerable, and less adaptable to extreme events (Marengo et al.,
2011, Debortoli et al., 2017). Adaptation is urgent in this context and directly depends
infrastructure and water resources (Marengo et al., 2021). The most common approach
statistical distribution of stationary annual maximums time series (Lang et al., 1999;
based on the dynamics of the natural system that float within an immutable envelope of
climatic variability (Milly et al., 2008). However, human interference may introduce
future (Field et al., 2012; Milly et al., 2008). Therefore, for technical solutions to be
effective and safe, they must be designed and operated considering climate change, not
depends on updated and accessible extreme events data. Frequently, the available data is
not generated in a format that managers and design engineers can easily use to mitigate
and adapt to future conditions (Markus et al., 2018). Thus, multidisciplinary studies
involving climate and engineering disciplines can play an essential role in assisting
Brazil (e.g., Porto de Carvalho et al., 2014; Silva Dias et al., 2013; Xavier et al., 2019,
Avila-Diaz et al., 2020b). More specific studies that address nonstationary extreme
precipitation series modeling are even more scarce (Assis et al., 2018; Pedron et al.,
2017; Zilli et al., 2017), and are focused on some municipalities in the south and
nonstationary extreme rainfall events for the entire national territory in any time period.
nonstationary conditions in Brazil and its projections for climate change scenarios. In
this sense, the study focused on the observational period during the last four decades
Projections (NEX-GDDP) and Brazilian National Institute for Space Research (INPE)
were used to assess nonstationary extreme rainfall for the 2020-2099 period. We expect
that the information provided in this research will be relevant to studies of impact,
sectors.
Annual extreme daily precipitation for the entire Brazilian territory (Figure 1) was
analyzed in this study. Much of the extreme rainfall occurs due to mesoscale systems
that can often be incorporated into synoptic systems that are influenced by large-scale
conditions, which in turn are also influenced by climatic variability (Cavalcanti, 2012).
Extreme precipitation in Brazil has been associated with anomalies both in the Pacific
and Atlantic Oceans, frontal systems, persistent systems such as the Intertropical
Convergence Zone (ITCZ) and South Atlantic Convergence Zone (SACZ), and
mesoscale convective systems (Lima et al., 2020; Cavalcanti, 2012; Luiz-Silva et al.,
2021). In addition, climate change increases the mean surface air temperature and sea
surface temperature, as well as changes the atmospheric water vapor content and
transport (Liu et al, 2020), with the potential to change the spatial patterns, intensity,
The maximum annual daily precipitation (mm.day-1) for each pixel of the study
area for the observational period (1980 to 2015) were obtained from the database of
Xavier et al. (Xavier et al., 2016), with the spatial resolution of 0.25°× 0.25° (latitude ×
longitude). This climate dataset was assumed as the reference because it is the most
complete gridded dataset, with 9,259 rain gauges to generate the daily fields of
precipitation over Brazil. In addition, this dataset has been used to analyze extreme
precipitation in many previous studies (Almagro et al., 2020; Avila-Diaz et al., 2020a,
(Thomson et al., 2011) and RCP8.5 (Riahi et al., 2011) scenarios from 2020 to 2099
We chose the climate model data based on Ávila-Diaz et al. (2020a), considering
the best response to reproduce the annual maximum 1-day precipitation (RX1day)
between downscaled models and gridded observational datasets in the historical period
(1980-2005). Therefore, the RX1day index followed the same definition and
Change Detection and Indices - ETCCDI (Seneviratne et al., 2021; Zhang et al., 2011).
Thus, we assessed downscaled climatic data from four climate models: the Hadley
Community Climate System Model, version 4 CCSM4 (Lawrence et al., 2012). This
study has focused on statistically and dynamically downscaled approaches (Table 1). In
this sense, HADGEM2-ES and MIROC5 models were downscaled using a dynamical
approach by the Weather Forecast and Climate Studies Center (CPTEC) of the National
Institute for Space Research (INPE) (Table 1). Furthermore, MRI-CGCM3 and CCSM4
GDDP) (Table 1) using the Global Meteorological Forcing Dataset (GMFD; Sheffield
et al., 2006) and the Bias-Correction Spatial Disaggregation algorithm (Thrasher et al.,
2012 ).
Model outputs were interpolated to the same spatial resolution of the Xavier et
al.’s data. Finally, the main results presented here were determined by the ensemble
average (individual model estimates are shown in the Supplementary Material, Figures
S2 and S5, as well as the uncertainty in simulated precipitation, Figures S10 to S13).
The generalized function of extreme values (GEV) provides a robust framework for
analyzing climatic extremes associated with return periods. This analysis structure
representing climatic variability over time (Katz et al., 2002; Ouarda and Charron,
2019). According to extreme values theory, annual maximums' behavior can generally
be represented by one of the three GEV families: Gumbel, Fréchet, and Reversed
family, through three parameters: location (µ), scale (σ), and shape (ξ) that governs the
tail behavior of the GEV distribution (Equations 1 and 2) (Coles, 2001). We adopted the
classic and stationary model, namely GEV0 (µ, σ, ξ), considering all parameters as
data, µ, σ and ξ are calculated for each model, dataset, and time-period in all pixels.
≠0
( |μ, , ) = − 1+ (1)
=0
( |μ, , ) = − − (2)
2.2.1. Nonstationary analysis
order to represent time variation in climate variables (e.g., Agilan and Umamahesh,
2017; Cheng et al., 2014; Almeida et al., 2016; El Adlouni et al., 2007; Tramblay et al.,
2013, 2012; Vu and Mishra, 2019; Westra et al., 2013; Zhou et al., 2012; Xavier at al.,
precipitation modeling, the GEV parameters must change over time (Coles, 2001). This
intricately linked to temporal variation has been widely adopted in recent studies
(Cheng et al., 2014; Sarhadi and Soulis, 2017; Vu and Mishra, 2019).
In the present study, to account for nonstationarity, the location (µ) and scale (σ)
parameters assume a time-dependent trend, and only the shape parameter (ξ) remains
constant, as the accurate estimation of this parameter is very challenging (Coles, 2001).
Thus, the nonstationary and time varying GEV function is defined as GEVt (µt, σt, ξ).
We assume the location and scale parameters as a linear function of time to account for
nonstationarity, and thus, µt and σt are given by Equations 3 and 4, respectively. Several
authors show that the linear variation of the parameters with time (e.g., Sarhadi and
Soulis, 2017; Vu and Mishra, 2019 Xavier at al., 2019; Liu et al., 2021) is a plausible
approximation of nonstationary extremes, and, with our data, the quadratic variation did
μ = + (3)
= + (4)
The parameters vector (θ) is estimated for each GEV model: three parameters
for GEV0, θ = (µ, σ, ξ); four parameters for GEV1, θ1 = ( , , , ), and five
and parameters are calculated for each model, dataset, and time-period in all pixels.
The maximum likelihood method (Smith, 1985) was used to calculate and obtain the
best estimates of all the statistical parameters of the GEV models. The method consists
function (ℓ) concerning each parameter. The solution of this system was determined
method, as widely used in similar studies (Vasiliades et al., 2015). The algorithm is
available in gev.fit function in the ismev free package running on R (Heffernan et al.,
2018). The initial guess of the parameters, when not set by the user, is calculated based
on the mean and variance of the precipitation data, as described in the Supplementary
The existence of trends in extreme precipitation was assessed using two tests (α=0,05):
the Mann-Kendall (Kendall, 1956; Mann, 1945) and the deviation (Coles, 2001) tests.
Mann-Kendall provides a preview of the existence and the sign of the trend in extreme
precipitation, highlighting the importance of nonstationary assumption. In addition, the
magnitude of the trend was determined by Sen’s slope estimator. The deviation test
compares the GEV models' validity based on the difference in their log of the maximum
likelihood. The deviation statistic (D, Equation 5) indicates if the nonstationary models
are better than the stationary model to represent the extreme precipitation (Coles, 2001;
El Adlouni et al., 2007; Tramblay et al., 2012; Xavier et al., 2019; Liu et al., 2021). The
2
asymptotic D distribution is given by the chi-square ( k) distribution with k degrees of
freedom. Here, k accounts for the difference in dimensionality (or complexity) of the
nonstationary and stationary models. Therefore, the extreme value distribution choice
= 2[ℓ ( )− ℓ ( )] (5)
Where ℓ (GEV1) and ℓ (GEV0) are the maximum likelihood logarithm of the
GEV1 and GEV0 models, respectively. Large D values indicate that the GEV1 is more
appropriate as it explains the data variation better than the GEV0, while the opposite
indicates that the increase in the model's complexity does not benefit the model's ability
to represent the data (Coles, 2001). Moreover, for nonstationary data, the use of
Equation 6 (comparing GEV1 and GEV2) identifies which nonstationary model better
represents extreme precipitation behavior. Therefore, for each pixel of the study area,
Equations 5 and 6 determined which GEV model was used to represent extreme
precipitation. Data that show adherence to GEV0 are stationary, and those that show
= 2[ℓ ( )− ℓ ( )] (6)
2.3.Calculation of extreme precipitation depths
Following previous studies (Agilan and Umamahesh, 2017; Cheng et al., 2014; Sarhadi
and Soulis, 2017; Tramblay et al., 2012), we calculated extreme precipitation depths
associated with the return periods (r) of 5, 10, 25, 50 and 100 years for the observational
periods (Equations 7 and 8 for stationary conditions and Equations 9 and 10 for
nonstationary conditions).
for ≠0
= μ − {1 − [− log(1 − )] } (7)
for =0
= μ− [− (1 − )] (8)
in the period p (observational, historical or future) associated with the return period r
(years), and µ, σ, ξ, are the location, scale, and shape parameters, respectively.
different GEV (and consequently different precipitation depths for each return period) is
calculated for each year of the analysis period, implying in an increasing (or decreasing)
frequency of extreme precipitation. Here we extend the notion of return period to the
events in r years is one´, where r is 5, 10, 20, 50, or 100 years. This has been suggested
as an efficient way to communicate risks for designing engineering projects when
nonstationarity has been identified and the notion of return period as in the stationary
case becomes unprecise (Cooley, 2013; Salas and Vogel, 2018; Yang 2019). In that
sense, to ensure infrastructure design safety, here we consider that the return levels
associated with each return period in the nonstationary case are the highest estimated
for ≠0
for =0
Extreme precipitation anomaly was determined for two scenarios (RCP4.5 and
RCP8.5), two time periods (2020-2059 and 2060-2099), and four climate models (Table
1). The analysis of projected precipitation data followed the same methodology as the
observational and historical periods: for each pixel, we identify the existence of trends,
estimate the GEV parameters for both stationary and nonstationary conditions and fit
3. Results
100 mm.day-1) over Brazil. For instance, Northern and Southern Brazil show the highest
(100 mm.day-1), and the interior northeastern region has the lowest (40 mm.day-1) value
(Figure 2-a).
b). According to the Mann-Kendall test, approximately 7.5% of the study area showed a
precipitation (dark blue areas in Figure 2-c). Positive trends are strong in the south and
northeastern coast. On the other hand, negative trends are strong in magnitude over the
central-western Amazon and central Brazil. Statistically, positive trends are as high as
1.2 mm.day-1year-1, which would translate into an increase in 43.2 mm.day-1 throughout
The deviation test indicated that 82.5% of the study area adhered to the
stationary model (GEV0). Furthermore, 16.7% of the area adhered to the model with the
location parameter as a function of time (GEV1), and only 0.8% to the model with both
location and scale parameters varying with time (GEV2). Thus, according to the
deviation test, nonstationary models (GEV1 and GEV2) represented 17.5% of the study
area (Figure 3). The deviation statistic showed good agreement with Mann-Kendall test
Extreme precipitation estimated under stationary conditions (Figure 4a-e) and estimated
deviance test, Figure 4f-j) show similar spatial distribution for each return period
( ), with differences that intensify as the recurrence time increases (Figure 4k-o).
Even though sparsely present in the country, this intensification is more accentuated in
the following states: Amazonas, Pará, Amapá, Sergipe, Alagoas, Mato Grosso, and São
Paulo. The difference between nonstationary and stationary models can be higher than
10 mm.
Even though most areas show increased precipitation depths under nonstationary
considered (mainly in Amazonas, Pará, Bahia, and Paraná, Figure 4-o), a pattern that
intensifies with the increase in return periods. Specifically, only 0%, 0.1%, 0.6%, 1.6%
and 2.6% of the area showed a reduction in precipitation depths under nonstationary
modeling, while 17.5%, 17.4%, 16.9%, 15.9% and 13.1% of the area showed an
increase in , for the return periods of 5, 10, 25, 50, and 100 years respectively
(Figure S1). This result strongly indicates that wrongly considering that the data is
Notably, most of the study area predicted future extreme precipitation depths are greater
than those calculated for the historical period, regardless of the time period, climate
change scenario, and associated return period (please see Figure 5-f to 5-o and Figure 6-
f to 6-o for the ensemble anomaly, and Figures S10 to S13 for the precipitation
magnitudes and uncertainty). In addition, for both periods and scenarios, there is an
S7). While in the historical period great part of the area showed extreme precipitation
ranging from 50 mm.day-1 to 80 mm.day-1, the climate change scenarios indicate that
extreme precipitation in the same areas is greater than 80 mm.day-1 in all return periods
greater frequency of specific return levels. For instance, extreme precipitation depths
that occurred once every 25 years in the historical period might occur more frequently.
In fact, they might occur every 5 years during 2020-2059 in the northeastern coast,
southern and southeastern regions of the country (Figures S10 to S13). Similarly,
according to the same projections, extreme precipitation depths that occurred once every
50 years in central Amazon, Rio de Janeiro, São Paulo, Acre, and Paraná might occur
every 10 years.
Furthermore, the greater the return period, the greater the increment in estimated
precipitation depths. In general, greater precipitation than historical period was found in
RCP 4.5 in 97.3%, 96.2%, 92.7%, 89% and 85.3% of the study area, while in RCP 8.5 it
reached 97.7%, 96.3%, 92.2%, 88% and 84.2%, for the return periods of 5, 10, 25, 50,
100 years, respectively, for 2020-2059 projections (Figure S8). Increases limited to 20
mm are prevalent (> 50% of the area) for r = 5, 10 and 25, while precipitation increases
precipitation is predicted to occur in the left and northern border of the country, as well
as from northwestern Amazon to southeastern Brazil (Figure 5 f-o). In the same period,
less intense precipitation is predicted in part of the Cerrado and the semi-arid region,
especially in RCP8.5 (Figure 5-o), where extreme precipitation depths decrease from 20
mm.day-1 in the five-year return period and as high as 150 mm.day-1 in the hundred-year
return period. The difference between the scenarios indicates that precipitation will be
more intense in RCP 8.5 in the majority of the area (Figure 5a-e).
2060-2099, even though they show an even greater increase in extreme precipitation
than historical period was found in RCP 4.5 in 98.6%, 97.4%, 93.7%, 90.1% and 86%
of the study area, while in RCP 8.5 it reached 99.5%, 99%, 96.6%, 93.1% and 95.3%,
for the return periods of 5, 10, 25, 50, 100 years, respectively (Figure S9). In addition,
increases limited to 20 mm are prevalent (> 50% of the area) for r = 5 and 10, while
precipitation increases greater than 40 mm are prevalent for r = 25, 50 and 100.
Finally, the most affected regions in climate change scenarios are the states of
Pernambuco, Amapá, Rio Grande do Sul, Rio de Janeiro, Espírito Santo, Minas Gerais,
and São Paulo. The last four states are the most populous states of Brazil.
4. Discussion
The results indicate that 17.5% of the study area presents nonstationarity in precipitation
extremes during the observational period. Therefore, the idea that impacts of extreme
precipitation might only occur in the future is incorrect and already potentially
occurring in many areas over the country. Furthermore, these changes were detected in
several areas with potential socioeconomic impacts, indicating that an assessment and
A wide range of statistical methods assists the design of hydraulic structures that
protect society and the environment from the impact of extreme precipitation events,
such as urban floods and landslides. Both the safety and financial cost of these
structures are related to the risk associated with the frequency of extreme precipitation,
determined by traditional methods that assume that the probability distribution functions
are stationary and governed by a single set of parameters (Milly et al., 2008; McPhillips
et al., 2018). However, our study indicates that stationary assumptions underestimate
extreme precipitation in 7.5% of the national territory, and infrastructure built in these
been to determine if traditional methods developed for stationary regimes can still be
are not based on the stationarity premise. Numerous recent studies have shown that
nonstationary GEV can be an efficient tool to explain the spatial and temporal evolution
and the dependencies between random variables and the climate (Cheng et al., 2014; El
Adlouni et al., 2007; Gonzalez-Alvarez et al., 2018; Katz et al., 2002; Xavier et al.,
expected to increase in at least 90% of the national territory. In the 5 and 10-year return
mm.day-1 in a few pixels. In 25, 50, and 100-year return periods increase from 20 to 60
mm.day-1, reaching more than 120 mm.day-1 for a few pixels. These results indicate
more intense and frequent disasters like urban floods and landslides, increasing social,
Even though CMIP5 models agree that Brazil will endure more variability in
these changes remains a challenge. Models have shown improvements in recent years,
but they still have difficulties in simulating precipitation patterns, mainly because the
However, studies indicate a future increase in ENOS variability (Cai et al., 2018),
changes in width and strength of ITCZ (Byrne et al., 2018); as well as changes in
position and intensity of SACZ (Zilli et al., 2021), all of which have potential to disturb
areas, increasing runoff. As more people live in densely populated urban areas, often
located in floodplains and low-lying coastal areas, their exposure to flood risks also
increases (Hammond et al., 2015). Urban drainage systems and infrastructure lifespan
are long (50 to 100 years), but the expected increase in runoff and the intensification of
precipitation extremes may prevent the drainage system from reaching the desired level
of service in the future (Alashan, 2018; Hammond et al., 2015; Mailhot and Duchesne,
2010; Willems et al., 2012; Xiong et al., 2019; Zahmatkesh et al., 2015; Zhou et al.,
2012). Such reality was evident in recent events of greater than average precipitation
that hit Minas Gerais, Rio de Janeiro, Espírito Santo, and São Paulo, causing many
impacts. The growing evidence of the increasing frequency of extremes urge the
development of new approaches to estimate risk and promote adaptation in urban areas,
and probabilistic models that incorporate nonstationarity are a significant part of this
The sole occurrence of extreme events does not necessarily lead to disasters but
resilience to extreme events and planning for recovery demands understanding risks
associated with extreme events and population vulnerabilities (McPhillips et al., 2018).
In addition, previous policy decisions on land use, zoning, and infrastructure affect the
level of risk exposure, susceptibility, and the adaptive capacity of the region to deal
with the impacts caused by extreme events (Cho and Chang, 2017).
More than 80% of the municipalities with very high vulnerability in Brazil are
located in the northern and northeastern regions, especially in the latter (Almeida et al.,
2016). The northeast region shows less exposure to natural risks than southern Brazil
but presents severe vulnerability conditions, with less adaptive and recovery capacity.
The south and southeastern regions have lower levels of vulnerability (Almeida et al.,
concentrate the largest areas of high or very high landslide susceptibility (IBGE,
2019a), as they have high population density located in areas of steep slopes. These
regions concentrate more than 56% of the Brazilian population (IBGE, 2019b).
precipitation in Brazil. The results of this work are subjected to many uncertainties.
First, modeling precipitation is a particularly challenging task, and models still diverge,
mainly in the change's magnitude. The differences in physical parameters between the
different models and the role of urbanization are still not well represented in the climate
models and are sources of uncertainty already identified by other studies (Cavalcanti et
al., 2017). Therefore, our estimates are based on a set of models from different
al., 2021). However, even in cases where the models do not agree on the magnitude,
they agree on the change signal, which indicates that the increase in the maximum
drive precipitation but may lose some local processes of convective or orographic order.
calculated for the historical period and affect the modeling process.
The difficulty of the calculation process and the lack of dialogue between the
scientific community and the technicians who apply these estimates inhibit the
difficulties can be surpassed by the development of software and digital platforms that
make this information clear and accessible, as well as training programs and courses to
transfer and disseminate the necessary knowledge. However, local government and
public managers make critical decisions that affect urban environments, and
communication with those actors is a crucial step to raise awareness and implement
such measures.
5. Conclusions
This study provides the first estimate of the frequency distribution of maximum daily
precipitation considering nonstationarity for the entire Brazilian territory. We show that
17.5% of the study area has nonstationary maximum daily precipitation in the historical
period and that neglecting nonstationarity, in most of the area, underestimates the
magnitudes of the daily maximum precipitation events for all return periods (e.g., 5, 10,
25, 50, and 100 years). Furthermore, considering nonstationarity, we found that
maximum daily precipitation might increase in at least 90% of the national territory in
the future, regardless of the climate scenario. Importantly, these changes are estimated
for the most populated regions of the country.
This scenario, along with accelerated urbanization, might explain the recent
increase in disasters related to urban floods and landslides over the country. In case
there are no adjustments, more intense and more frequent urban floods, with the
increase in social, environmental, and economic risks, might occur. It suggests the
urgent need to prepare and adapt urban environments to the current and future extreme
To this end, large-scale analysis like the presented in this work should be refined
and include local specificities that may increase the accuracy of the estimated frequency
simulation in finer resolutions (at least 1km), evaluating the steepness of local reliefs
and population occupation of high disaster risk areas. However, the greatest challenge to
frequency and intensity determine urban infrastructures' risk and economic cost, and the
Finally, there is still a long way to adapt urban environments to climate change,
and local managers and decision-makers must be aware that this is an urgent matter, and
the lack of action exposes society to an imminent risk and might cause the loss of
The authors thank the graduate program in Applied Meteorology at the Federal
University of Viçosa. This study was supported by The National Council for Scientific
and Technological Development (CNPq) and the Coordination of Training of Higher
Education Personnel (CAPES). AAD also thanks the financial support received from
CNPq postdoctoral Programme. The authors are thankful for the downscaled climate
simulations generated by CPTEC/INPE and made available on the PROJETA platform,
and we thank NEX-GDDP dataset, prepared by the Climate Analytics Group and NASA
Ames Research Center using the NASA Earth Exchange, and distributed by the NASA
Center for Climate Simulation (NCCS).
References
Agilan, V., Umamahesh, N. V., 2017. What are the best covariates for developing non-
stationary rainfall Intensity-Duration-Frequency relationship? Advances in Water
Resources, 101, 11–22. https://doi.org/10.1016/j.advwatres.2016.12.016
Alexander, L. V., et al., 2006. Global observed changes in daily climate extremes of
temperature and precipitation. Journal of Geophysical Research Atmospheres,
111(5), 1–22. https://doi.org/10.1029/2005JD006290
Almeida, L. Q., Welle, T., Birkmann, J., 2016. Disaster risk indicators in Brazil: A
proposal based on the world risk index. International Journal of Disaster Risk
Reduction, 17, 251–272. https://doi.org/10.1016/j.ijdrr.2016.04.007
Alves, L. M., et al., 2021. Assessment of rainfall variability and future change in Brazil
across multiple timescales. International Journal of Climatology, 41(S1), E1875–
E1888. https://doi.org/10.1002/joc.6818
Assis, L. C., et al., 2018. A model-based site selection approach associated with
regional frequency analysis for modeling extreme rainfall depths in Minas Gerais
state, Southeast Brazil. Stochastic Environmental Research and Risk Assessment,
32(2), 469–484. https://doi.org/10.1007/s00477-017-1481-1
Ávila, A., et al., 2016. Recent precipitation trends, flash floods and landslides in
southern Brazil. Environmental Research Letters, 11(11).
https://doi.org/10.1088/1748-9326/11/11/114029
Avila-Diaz, A., et al., 2020b. Assessing current and future trends of climate extremes
across Brazil based on reanalyses and earth system model projections. Climate
Dynamics, 55(5–6), 1403–1426. https://doi.org/10.1007/s00382-020-05333-z
Byrne, M. P., et al., 2018. Response of the Intertropical Convergence Zone to Climate
Change: Location, Width, and Strength. Current Climate Change Reports, 4(4),
355–370. https://doi.org/10.1007/s40641-018-0110-5
Cai, W., et al., 2018. Increased variability of eastern Pacific El Niño under greenhouse
warming. Nature, 564(7735), 201–206. https://doi.org/10.1038/s41586-018-0776-9
Cavalcanti, I. F. A., 2012. Large scale and synoptic features associated with extreme
precipitation over South America: A review and case studies for the first decade of
the 21st century. Atmospheric Research, 118, 27–40.
https://doi.org/10.1016/j.atmosres.2012.06.012
Chen, D., M., et al., 2021: Framing, Context, and Methods. In Climate Change 2021:
The Physical Science Basis. Contribution of Working Group I to the Sixth
Assessment Report of the Intergovernmental Panel on Climate Change [Masson-
Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y.
Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R.
Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)].
Cambridge University Press. In Press.
Cheng, L., et al., 2014. Non-stationary extreme value analysis in a changing climate.
Climatic Change, 127(2), 353–369. https://doi.org/10.1007/s10584-014-1254-5
Cho, S. Y., Chang, H., 2017. Recent research approaches to urban flood vulnerability,
2006–2016. Natural Hazards, 88(1), 633–649. https://doi.org/10.1007/s11069-017-
2869-4
Chou, S. C., et al., 2014a. Assessment of Climate Change over South America under
RCP 4.5 and 8.5 Downscaling Scenarios. American Journal of Climate Change,
03(05), 512–527. https://doi.org/10.4236/ajcc.2014.35043
Chou, S. C., et al., 2014b. Evaluation of the Eta Simulations Nested in Three Global
Climate Models. American Journal of Climate Change, 03(05), 438–454.
https://doi.org/10.4236/ajcc.2014.35039
Coles, S., 2015. An introduction to extreme values. In The effects of brief mindfulness
intervention on acute pain experience: An examination of individual difference
(Vol. 1).
Cooley, D., 2013. Return Periods and Return Levels Under Climate Change. Extremes
in a Changing Climate, 25. 97-114 https://doi.org/10.1007/978-94-007-4479-0_4
Costa, R. L., et al., 2020. Analysis of climate extremes indices over northeast Brazil
from 1961 to 2014. Weather and Climate Extremes, 28.
https://doi.org/10.1016/j.wace.2020.100254
de Loyola Hummell, B. M., Cutter, S. L., & Emrich, C. T., 2016. Social Vulnerability to
Natural Hazards in Brazil. International Journal of Disaster Risk Science, 7(2),
111–122. https://doi.org/10.1007/s13753-016-0090-9
Debortoli, N. S., Camarinha, P. I. M., Marengo, J. A., & Rodrigues, R. R., 2017. An
index of Brazil’s vulnerability to expected increases in natural flash flooding and
landslide disasters in the context of climate change. Natural Hazards, 86(2), 557–
582. https://doi.org/10.1007/s11069-016-2705-2
Donat, M. G., et al., 2016. More extreme precipitation in the worldâ €TM s dry and wet
regions. Nature Climate Change, 6(5), 508–513.
https://doi.org/10.1038/nclimate2941
El Adlouni, S., et al., (2007). Generalized maximum likelihood estimators for the
nonstationary generalized extreme value model. Water Resources Research, 43(3),
1–13. https://doi.org/10.1029/2005WR004545
Eyring, V., et al., 2021. Chapter 3: Human influence on the climate system.
In: AR6 Climate Change 2021: The Physical Science Basis. IPCC, 207.
04-202. https://doi.org/10.1002/2014GL06106
Field, C.B., et al., eds., 2012. Managing the risks of extreme events and disasters to
advance climate change adaptation. Special Report of Working Groups I and II of
the Intergovernmental Panel on Climate Change (IPCC). Cambridge: Cambridge
University Press.
Jenkinson, A. F., 1955. The frequency distribution of the annual maximum (or
minimum) values of meteorological elements. Quarterly Journal of the Royal
Meteorological Society, 81(348), 158–171.
https://doi.org/10.1002/qj.49708134804
Katz, R. W., Parlange, M. B., Naveau, P., 2002. Statistics of extremes in hydrology.
Advances in Water Resources, 25(8–12), 1287–1304.
https://doi.org/10.1016/S0309-1708(02)00056-8
Lang, M., Ouarda, T. B. M. J., Bobée, B., 1999. Towards operational guidelines for
over-threshold modeling. Journal of Hydrology, 225(3–4), 103–117.
https://doi.org/10.1016/S0022-1694(99)00167-5
Lawrence, D. M., et al., 2012. The CCSM4 land simulation, 1850-2005: Assessment of
surface climate and new capabilities. Journal of Climate, 25(7), 2240–2260.
https://doi.org/10.1175/JCLI-D-11-00103.1
Lehmann, J., Coumou, D., Frieler, K., 2015. Increased record-breaking precipitation
events under global warming. Climatic Change, 132(4), 501–515.
https://doi.org/10.1007/s10584-015-1434-y
Lehtonen, I., Jylhä, K., 2019. Tendency towards a more extreme precipitation climate in
the Coupled Model Intercomparison Project Phase 5 models. Atmospheric Science
Letters, 20(5), 1–7. https://doi.org/10.1002/asl.895
Lucas, E. W. M., et al., 2021. Trends in climate extreme indices assessed in the Xingu
river basin - Brazilian Amazon. Weather and Climate Extremes, 31(January),
100306. https://doi.org/10.1016/j.wace.2021.100306
Liu, B., et al., 2020. Global atmospheric moisture transport associated with precipitation
extremes: Mechanisms and climate change impacts. Wiley Interdisciplinary
Reviews: Water, 7(2), 1–25. https://doi.org/10.1002/wat2.1412
Lyra, A., et al., 2018. Climate change projections over three metropolitan regions in
Southeast Brazil using the non-hydrostatic Eta regional climate model at 5-km
resolution. Theoretical and Applied Climatology, 132(1–2), 663–682.
https://doi.org/10.1007/s00704-017-2067-z
Mailhot, A., Duchesne, S., 2010. Design Criteria of Urban Drainage Infrastructures
under Climate Change. Journal of Water Resources Planning and Management,
136(2), 201–208. https://doi.org/10.1061/(asce)wr.1943-5452.0000023
Mann, H. B., 1945. Non-Parametric Test Against Trend. Econometrica, 13(3), 245–259.
http://www.economist.com/node/18330371?story%7B_%7Did=18330371
Marengo, J. A., et al., 2011. Recursos hídricos em regiões áridas e semiáridas. Recursos
hídricos em regiões áridas e semiáridas.
http://plutao.sid.inpe.br/col/dpi.inpe.br/plutao/2011/06.11.02.16/doc/Marengo_Var
iabilidade.pdf?languagebutton=en
Markus, M., et al., 2018. Communicating the Impacts of Projected Climate Change on
Heavy Rainfall Using a Weighted Ensemble Approach. Journal of Hydrologic
Engineering, 23(4), 04018004. https://doi.org/10.1061/(asce)he.1943-
5584.0001614
Milly, P. C. D., et al., 2008. Climate change: Stationarity is dead: Whither water
management? Science, 319(5863), 573–574.
https://doi.org/10.1126/science.1151915
O’Gorman, P. A., & Schneider, T. 2009. The physical basis for increases in
precipitation extremes in simulations of 21st-century climate change. Proceedings
of the National Academy of Sciences of the United States of America, 106(35),
14773–14777. https://doi.org/10.1073/pnas.0907610106
Parey, S., Hoang, T.T.H., Dacunha-Castelle, D., 2010. Different ways to compute
temperature return levels in the climate change context. Environmetrics, 21, 698–
718. https://doi.org/10.1002/env.1060
Porto de Carvalho, J. R., et al., 2014. Annual maximum daily rainfall trends in the
midwest, southeast and southern Brazil in the last 71 years. Weather and Climate
Extremes, 5(1), 7–15. https://doi.org/10.1016/j.wace.2014.10.001
Riahi, K., et al., 2011. RCP 8.5-A scenario of comparatively high greenhouse gas
emissions. Climatic Change, 109(1), 33–57. https://doi.org/10.1007/s10584-011-
0149-y
Salas, J. D., Obeysekera, J., Vogel, R. M., 2018. Techniques for assessing water
infrastructure for nonstationary extreme events: a review. Hydrological Sciences
Journal, 63(3), 325–352. https://doi.org/10.1080/02626667.2018.1426858
Seneviratne, S. I., et al., 2021. Weather and Climate Extreme Events in a Changing
Climate. In: Climate Change 2021: The Physical Science Basis. Contribution of
Working Group I to the Sixth Assessment Report of the Intergovernmental Panel
on Climate Change. In P. Z. Masson-Delmotte, V., E. A. Pirani, S. L. Connors, C.
Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M. I. Gomis, M. Huang, K.
Leitzell, & R. Y. and B. Z. Lonnoy, J. B. R. Matthews, T. K.
Sheffield J., Goteti G., Wood E.F., 2006. Development of a 50-year high-resolution
global dataset of meteorological forcings for land surface modeling. Journal Of
Climate.19(13), 3088-3111. doi:10.1175/JCLI3790.1
Silva Dias, M. A. F., et al., 2013. Changes in extreme daily rainfall for São Paulo,
Brazil. Climatic Change, 116(3–4), 705–722. https://doi.org/10.1007/s10584-012-
0504-7
Thomson, A. M., et al., 2011. RCP4.5: A pathway for stabilization of radiative forcing
by 2100. Climatic Change, 109(1), 77–94. https://doi.org/10.1007/s10584-011-
0151-4
Thrasher, B., et al., 2012. Technical Note: Bias correcting climate model simulated
daily temperature extremes with quantile mapping. Hydrology and Earth System
Sciences, 16(9), 3309–3314. https://doi.org/10.5194/hess-16-3309-2012
Tramblay, Y., Neppel, L., Carreau, J., Najib, K., 2013. Analyse fréquentielle non-
stationnaire des pluies extrêmes dans le Sud de la France. Hydrological Sciences
Journal, 58(2), 280–294. https://doi.org/10.1080/02626667.2012.754988
Vasiliades, L., Galiatsatou, P., Loukas, A., 2015. Nonstationary Frequency Analysis of
Annual Maximum Rainfall Using Climate Covariates. Water Resources
Management, 29(2), 339–358. https://doi.org/10.1007/s11269-014-0761-5
Vu, T. M., Mishra, A. K., 2019. Nonstationary frequency analysis of the recent extreme
precipitation events in the United States. Journal of Hydrology, 575(May), 999–
1010. https://doi.org/10.1016/j.jhydrol.2019.05.090
Watanabe, M., et al., 2010. Improved climate simulation by MIROC5: Mean states,
variability, and climate sensitivity. Journal of Climate, 23(23), 6312–6335.
https://doi.org/10.1175/2010JCLI3679.1
Westra, S., Alexander, L. V., Zwiers, F. W., 2013. Global increasing trends in annual
maximum daily precipitation. Journal of Climate, 26(11), 3904–3918.
https://doi.org/10.1175/JCLI-D-12-00502.1
Willems, P., Arnbjerg-Nielsen, K., Olsson, J., Nguyen, V. T. V., 2012. Climate change
impact assessment on urban rainfall extremes and urban drainage: Methods and
shortcomings. Atmospheric Research, 103, 106–118.
https://doi.org/10.1016/j.atmosres.2011.04.003
Xavier, A. C., King, C. W., Scanlon, B. R., 2016. Daily gridded meteorological
variables in Brazil (1980–2013). International Journal of Climatology, 36(6),
2644–2659. https://doi.org/10.1002/joc.4518
Xiong, L., et al., 2019. Impacts of Climate Change on Urban Extreme Rainfall and
Drainage Infrastructure Performance: A Case Study in Wuhan City, China.
Irrigation and Drainage, 68(2), 152–164. https://doi.org/10.1002/ird.2316
Yang L., Li J., Sun H., Guo Y., Engel B.A. 2019. Calculation of nonstationary flood
return period considering historical extraordinary flood events. Journal Flood Risk
Management. 12(2), 1-10. doi:10.1111/jfr3.12463
Yukimoto, S., et al., 2012. A new global climate model of the Meteorological Research
Institute: MRI-CGCM3: -Model description and basic performance-. Journal of the
Meteorological Society of Japan, 90(A), 23–64. https://doi.org/10.2151/jmsj.2012-
A02
Zhang, X., et al., 2011. Indices for monitoring changes in extremes based on daily
temperature and precipitation data. Wiley Interdisciplinary Reviews: Climate
Change, 2(6), 851–870. https://doi.org/10.1002/wcc.147
Zhou, Q., et al., 2012. Framework for economic pluvial flood risk assessment
considering climate change effects and adaptation benefits. Journal of Hydrology,
414–415, 539–549. https://doi.org/10.1016/j.jhydrol.2011.11.031
Zilli, M. T., et al., 2017. A comprehensive analysis of trends in extreme precipitation
over southeastern coast of Brazil. International Journal of Climatology, 37(5),
2269–2279. https://doi.org/10.1002/joc.4840
TABLES
Figure 2. Extreme precipitation during the observational period (1985-2015). (a) average
maximum annual precipitation (AMAP, mm.day-1), (b) extreme precipitation trend (EPT,
mm.day-1.year-1) determined by Sen’s slope estimator and (c) p-value for Mann-Kendall test
(dark blue areas indicate statistically significant trends, α=0.05).
Figure 3. Spatial distribution of the adherence to the different GEV models according to the
deviation test for the historical period (1985-2015).
Figure 4. Historical maximum precipitation depths associated with the return periods of 5, 10,
25, 50, and 100 years ( , , , , ), in mm.day-1 (a) to (e) estimated with the
stationary model (GEV0), (f) to (j) with the model indicated by the deviation test (GEV0, GEV1,
GEV2) and (k) to (o) the difference between them.
Figure 5. Future maximum precipitation depth anomaly, from 2020 to 2059, associated
with the return periods of 5, 10, 25, 50, and 100 years ( , , , , ),
in mm.day-1. (a) to (e) between RCP 8.5 and RCP 4.5, (f) to (j) between RCP 4.5 and
the historical period, and (k) to (o) between RCP 8.5 and the historical period.
Figure 6. Future maximum precipitation depth anomaly, from 2060 to 2099, associated with the
return periods of 5, 10, 25, 50, and 100 years ( , , , , ), in mm.day-1. (a)
to (e) between RCP 8.5 and RCP 4.5, (f) to (j) between RCP 4.5 and the historical period, and
(k) to (o) between RCP 8.5 and the historical period.