133
A FIRST APPROACH TO CONSIDER THE
INFORMATION OF THE MADDEN-JULIAN
OSCILLATION IN THE OPERATION OF THE
ELECTRICAL SYSTEM OF URUGUAY
PRIMEROS PASOS PARA CONSIDERAR LA INFORMACIÓN DE LA
OSCILACIÓN MADDEN-JULIAN EN LA OPERACIÓN DEL SISTEMA
ELÉCTRICO DE URUGUAY
Matilde Ungerovich1, Ruben Chaer2, Felipe Palacio3, Guillermo Flieller4
Recibido: 19/11/2024 y Aceptado: 13/10/2025
1.- matildeungerovich@gmail.com
2.- rchaer@adme.com.uy
3.- fpalacio@adme.com.uy
4.- gieller@adme.com.uy
Sección otros artículos
134
135
La Oscilación Madden-Julian (MJO) es una perturbación intraestacional (30-90 días) en la atmósfera
tropical que inuye en conguraciones climáticas en distintas regiones. Por ejemplo, en el sudeste de
América del Sur la MJO afecta las precipitaciones, especialmente durante verano austral, con fases
que favorecen lluvias extremas en Uruguay y el sur de Brasil, inuyendo en el caudal de embalses
hidroeléctricos de Uruguay. La importancia de la MJO radica en que puede predecirse con hasta cinco
semanas de antelación, permitiendo anticipar sus efectos en distintas regiones. En este estudio se
compara la programación energética óptima del país considerando y sin considerar los efectos de la
oscilación. Se simulan posibles realizaciones estocásticas de las condiciones futuras y se calcula la
programación energética óptima. En la mitad de los casos se considera la información de MJO y en
la otra mitad no. Los resultados indican que incluir la información histórica de MJO afecta el consumo
de gasoil. En particular, cuando se considera la oscilación, la fase Niño muestra un comportamiento
menos extremo y con menor variabilidad que cuando no se considera.
The Madden-Julian Oscillation (MJO) is an intraseasonal oscillation (30-90 days) in the tropical
atmosphere that inuences climate patterns in various regions. For example, in Southeastern South
America, the MJO impacts rainfall, especially during the austral summer, with phases that favor extreme
rainfall in Uruguay and southern Brazil, aecting the inows to Uruguay’s hydroelectric reservoirs. The
importance of the MJO lies in its predictability, which extends up to ve weeks in advance, allowing
for the anticipation of its eects. This study compares the country’s optimal energy programming,
considering and not considering the eects of MJO. Possible stochastic realizations are simulated,
and the optimal energy programming is calculated. In half of the cases, MJO information is considered,
while in the other half, it is not. Results indicate that including historical MJO information aects diesel
consumption. In particular, when the oscillation is taken into account, the El Niño phase exhibits less
extreme behavior and lower variability than when it is not considered.
PALABRAS CLAVE: Oscillación de Madden-Julian, Uruguay, Energía hidráulica, ENSO, SIMSEE
KEYWORDS: Madden-Julian Oscillation, Uruguay, hydropower, ENSO, SIMSEE
Resumen
Abstract
Sección otros artículos
136
1. INTRODUCTION
The Madden-Julian Oscillation or MJO (Zhang,
2005) is an intraseasonal disturbance (30-90
days) in the tropical atmosphere that signi cantly
impacts global climate conditions. It is an
eastwardly propagating cell characterized by an
enhanced and a suppressed convective region.
The evolution of MJO is most typically represented
by the Real-time Multivariate MJO index (RMM,
Wheeler and Hendon, 2004). It is a numerical
index that, considering 850 and 200 hPa zonal
wind and outgoing longwave radiation, quanti es
The importance of the MJO lies in the fact that it
can be predicted with a lead time of ve weeks
(Kim et. al., 2018), allowing for the prediction of
its worldwide e ects. In particular, the e ects of
MJO in Southeastern South America have been
analyzed in many publications. For example,
(Alvarez, et. al., 2016) found that during austral
summer phases 3, 4, and 5 favor simultaneous
weekly rainfall in the upper tercile in Uruguay and
southern Brazil (a region that in uences the ow
the intensity and phase of the oscillation. The RMM
index is composed of two principal components,
RMM1 and RMM2, which together de ne a two-
dimensional phase space. The angle of the vector
(RMM1, RMM2) indicates the location that has an
associated phase (from 1 to 8), and its magnitude
re ects the strength (amplitude) of the convective
signal. Figure 1 shows an example for the period
27 May-5 July, with a weak MJO in May and
July, and a higher intensity of phases 6, 7, and 8
between 4 and 12 June.
Figure 1. Example of Madden-Julian Oscillation vector (RMM1, RMM2) diagram from 27 May 2025 until 5 July 2025.
Taken from http://www.bom.gov.au/
in the most important Uruguayan hydroelectric
power plant) while in austral autumn phases, 4,
5 and 6 (8) are associated to enhanced (reduced)
precipitation; in spring phases 4 and 5 are related
to upper tercile. In winter, the relationship is less
important. Additionally, in Ungerovich et. al. (2021),
the authors conclude that the persistence of the
MJO for more than ve days in phases 4 and 5
during austral spring is a precursor to extreme
rainfall events in southern Uruguay.
ENERLAC
137
2. METHODS
The Electric System Operation and Expansion
Simulation (SimSEE, Chaer, 2008) is a modeling
tool developed in Uruguay to analyze the behavior
of electric power systems, particularly those
combining hydroelectric and thermal generation.
It enables simulation of system operation under
varying hydrological and demand conditions
The Uruguayan precipitation regime imposes
signicant variability in the annual energy available
from this source. The annual generation of the
hydroelectric subsystem ranges from 3,300 to
9,300 GWh (BEN, 2023). The largest reservoir,
located on the Río Negro river, can store enough
energy to operate at full capacity (596 MW) for up
to 135 days when full. It feeds a chain of three
power plants (Chaer, 2008). Additionally, the
binational Uruguayan-Argentinian Salto Grande
hydroelectric plant on the Uruguay River has
an installed capacity of 1800 MW, half of which
corresponds to Uruguay, and a storage capacity
of ve days. National demand is about 1,300 MW
(annual average), with peak values of about 2,200
MW and minimum values of around 700 MW.
The sum of wind (1,550 MW) and solar (220 MW)
installed capacity exceeds the daily peak demand
on 70% of the days of the year. For instance, in
2023, the Uruguayan power system supplied a
national demand of 11,472 GWh plus an export of
244 GWh. This energy was fullled by 39% wind,
3% solar, 9% biomass, 28% hydroelectric, 8%
thermal, and 12% imports (ADME, 2025).
The main challenge for the system’s optimal
operation is the economic valuation of water
resources from the three main reservoirs. The
programming of the National Interconnected
System (SIN) is carried out by the Electricity
Market Administration (ADME). To achieve this, it
utilizes two automatic power dispatch programs:
Vates_MP and Vates_CP (ADME, 2023). They are
constantly assimilating information on the state of
the SIN, the forecasts of the surface temperature
anomaly of the Pacic Ocean in the El Niño region,
ow rates of contributions to the lakes, wind
speed, solar radiation, and temperature.
Chaer et. al. (2010) provide the foundation
for incorporating El Niño-Southern Oscillation
(ENSO) forecasts into Uruguay’s energy dispatch
programming. Although the initial concept was
developed in 2010, it was formally published in
2015 (Maciel et. al., 2015), providing a detailed
approach to incorporating ENSO-related climate
signals to optimize Uruguay’s energy system
operation. The paper focused on integrating
ENSO forecasts into the stochastic modeling of
streamow, aiming to reduce operational costs
by improving the management of hydroelectric
resources, which are highly dependent on
interannual climatic variations. This approach
enables the system to anticipate periods of
drought or excessive rainfall better, adjusting
energy dispatch accordingly to ensure a more
ecient and cost-eective operation.
This paper examines the incorporation of the MJO
as an additional tool in the dispatch framework,
serving as a complementary approach to enhance
power systems operation under uncertainty.
Specically, the objective of this study is to evaluate
the impact of incorporating MJO information into
stochastic simulations used for medium-term
energy planning, with a focus on its eect on
diesel consumption under dierent ENSO phases.
2.1 Simulation model: SimSEE
and is widely used for both long-term planning
and short-term operational studies. SimSEE
operates with Correlations in Gaussian Space
using Histograms (CEGH, Chaer et. al., 2011),
a stochastic modeling framework that generates
synthetic time series while preserving the key
statistical features of historical data.
Sección otros artículos
138
For this study, two cases were considered:
1. MI (More Informed): Using historical
information from both the MJO and ENSO
phases. This means that the CEGH will be
based on historical information regarding
the MJO, ENSO, and water ow rates in the
hydroelectric dams.
2. LI (Less Informed): Only considering the
historical information from ENSO phases
and water ow data. This means that the
CEGH will be based on historical information
about ENSO and the water ow rates in the
hydroelectric dams.
The study focused on the impact of the MJO
on fuel consumption in the electrical system,
particularly during the austral summer (December-
January-February). We considered a closed
system without imports or exports of energy.
Then, Uruguay’s energy system is composed of
both renewable and thermal energy sources. In
that scenario, considering the amount of water
available today and the amount that will be available
during the following days, a decision is made on
when and how much thermal energy will be used.
To understand the MJO e ect, we will analyze
the amount of diesel that thermal machines will
need over the next 90 days, considering both with
and without the historical information of MJO (MI:
more informed and LI: less informed, respectively).
Speci cally, we ran ve sets of 3000 stochastic
2.2 Variables and scenarios
simulations using di erent initial random seeds
(S1-S5) and analyzed thermal energy dispatch
decisions. The idea behind the ve sets is to
make the results more robust than with an only
set. Then, we analyzed how diesel consumption
varies in the MI and LI simulations under the three
phases of ENSO.
The CEGH models were trained to calculate
incoming water ows to the lakes associated
with hydroelectric plants and then determine the
amount of diesel to be purchased to meet the
thermal energy needs. To estimate the value of the
information provided by the history of the MJO,
statistical measures associated with the expected
value of the operation’s cost over the next 90 days
were calculated.
3. RESULTS
For the purpose of assessing the e ect of MJO in
the availability of hydroelectric resources we shall
de ne the Incoming Hydroelectric Energy (IHE) as
the sum of the product of the in ow to each dam
and the energetic coe cient given the height of the
3.1 MJO correlation with hydropower
lake and the downstream river. IHE is presented in
equation 1, where ρ, g, Q, and h correspond to
water density, gravitational acceleration, ow rate
and height, respectively.
Equation 1- Incoming hydroelectric energy
ENERLAC
139
Figures 2 and 3 show the results of the average
of the ve sets of 3,000 simulations of cumulative
90-day diesel consumption under the three ENSO
phases: La Niña, Neutral, and El Niño. The results
correspond to the average of the ve initial seeds,
and the simulations have been sorted in ascending
order of diesel usage to visualize the distribution
across simulations.
The gures show that during the highest diesel
demand periods (characterized by less rainfall) in
both cases (MI and LI), La Niña corresponds to
higher demand than El Niño. On the other hand, in
the more rainy simulations (lower diesel demand),
for LI cases, the demand is almost independent of
ENSO, while for MI, El Niño seems to imply more
diesel consumption than La Niña. However, the
dierence is less than 0.1 hm3, and as SimSEE
also takes into account the economic aspect, it is
not safe to make conclusions about the dierence
in rainfall.
Additionally, in MI, the variation in consumption
between ENSO phases is more pronounced
than in LI, with the most signicant dierences
observed in the intermediate and driest periods
(characterized by medium and high diesel
consumption). These results are also shown in
Table 1, which displays diesel consumption for
the three ENSO phases. The data is presented for
percentiles 10, 50, and 90 of the average of the
ve sets of 3000 simulations. Diesel consumption
values are shown for both cases: LI and MI.
In addition, gures 4 and 5 present a more detailed
analysis of 90-day diesel consumption using
simulations initialized with the ve dierent random
seeds for LI and MI, respectively. Both gures show
3.2 Impact of MJO on diesel consumption
Fig. 2 shows the iN34 index, RMM1, and RMM2
correlations with IHE. The rst thing to observe
is that the iN34 index presents correlations with
the IHE that are three times higher than those
observed with the components of the MJO.
In the operation, the forecast for the following 10
days is taken from meteorological forecasts and
assimilated into the stochastic models to schedule
the energy dispatch. The possible contribution of
new information from the MJO is then in the time
horizon after those rst ten days. As shown in the
gure, the RMM2 component exhibits a signicant
correlation with the IHE 15 days in advance.
the 10th, 50th, and 90th percentiles, as well as the
standard deviation of diesel consumption across
ENSO phases. Comparing Figures 4 and 5 reveals
that when MJO is considered, diesel consumption
during El Niño increases in both the lower and
upper extremes. This means that the 10th and
especially the 90th percentiles are higher than in
the case without MJO, suggesting that both dry
and wet El Niño scenarios result in greater diesel
use when MJO is taken into account. Specically,
the wettest El Niño years (10th percentile) require
more diesel than when MJO is ignored. Likewise,
but to a lesser extent, the driest El Niño years (90th
percentile) become even drier, intensifying diesel
needs. Moreover, the standard deviation is much
lower in simulations that include MJO, indicating
that diesel consumption during El Niño becomes
more consistent and predictable. During La Niña
or Neutral years, the dierences between including
and excluding the MJO are less signicant, both in
terms of percentiles and variability.
Sección otros artículos
140
Figure 2. Correlations of iN34 and MJO with IHE
Figure 3. Cumulative 90-day diesel consumption from 3,000 simulations without taking into account MJO’s historical
information, under the three ENSO phases (La Niña, Neutral, and El Niño), using the average of the 5 seeds. Simulations
are ordered from lowest to highest consumption to illustrate the distribution of outcomes. The left panel shows all the
simulations, the middle one shows the 700 ones with the highest diesel consumption and the right one shows
the lowest 700.
Source: Own elaboration
Source: Own elaboration
ENERLAC
141
Table 1. Diesel consumption values for percentiles 10, 50 and 90 of the average of the ve sets of 3000 simulations under
the scenarios of Niña, Neutral, and Niño. The values are shown for both conditions: LI and MI
Figure 4. Similar to gure 2 but MI
Source: Own elaboration
Source: Own elaboration
Sección otros artículos
142
Figure 5. Diesel consumption over 90 days for simulations that do not take into account MJO phases (LI) and are initialized
with 5 di erent seeds (S1-S5). The 10th, 50th, and 90th percentiles, along with the standard deviation, are shown
Source: Own elaboration
ENERLAC
143
Figure 6. As gure 4 but considering MJO phases (MI)
Source: Own elaboration
Sección otros artículos
144
4. CONCLUSIONS
5. DISCLAIMER
The research presented in this paper represents
the rst attempt to incorporate Madden-Julian
Oscillation (MJO) information into Uruguay’s
energy dispatch programming. Based on the
results obtained, some conclusions can be drawn.
The study shows that the iN34 index (representing
ENSO) has a stronger and more persistent
correlation with incoming hydraulic energy than
the MJO components. However, the RMM2
component of the MJO shows a relevant
correlation 15 days in advance.
For dry and intermediate seasons, independent
of MJO considerations, the analysis highlights
seasonal dierences in diesel consumption, with
the drier La Niña phase requiring more diesel
due to reduced rainfall and the El Niño phase
requiring less. However, during rainy seasons,
the relationship between consumption and
ENSO phase diers for MI and LI. Also, including
MJO information makes these dierences more
The content of this article is entirely the responsibility of its authors, and does not necessarily reect the
position of the institutions of which they are part of.
pronounced in the extreme values. Finally, it is
shown that considering MJO during the El Niño
phase results in higher percentiles for the highest
diesel consumption, indicating greater need and
suggesting lower rainfall. Additionally, the standard
deviation is much lower. On the other hand, during
La Niña or neutral years, the eects of MJO are
less signicant.
Previous analyses have demonstrated that the
MJO inuences rainfall in the region and that
SimSEE can accurately reproduce the oscillation.
Therefore, the fact that diesel consumption
changes when MJO information is incorporated
suggests that the results may be improved.
Incorporating MJO could improve the decisions
made by ADME, allowing for more accurate
planning. Incorporating MJO into energy dispatch
models not only improves predictive consistency
but also enhances resilience in energy planning
under climate variability.
ENERLAC
145
6. REFERENCES
Administración del Mercado Eléctrico (ADME). (2023). *Optimum operation of electric power generation resources*.
https://adme.com.uy/db-docs/Docs_secciones/nid_1457/IntoOperacionOptimaGen_ADME_feb2023_en.pdf
ADME. (2025). Datos de operación. Recuperado de https://adme.com.uy/controlpanel.php
Alvarez, M. S., Vera, C. S., Kiladis, G. N., & Liebmann, B. (2016). Inuence of the Madden Julian Oscillation on
precipitation and surface air temperature in South America. Climate Dynamics, 46, 245-262.
BEN (2023). Balance Energético Nacional 2023. Ministerio de Industria, Energía y Minería (MIEM), Uruguay.
Disponible en https://ben.miem.gub.uy/descargas/1balance/1-1-Book-BEN2023.pdf
Chaer, R. (2008). Simulación de sistemas de energía eléctrica.
Chaer, R., Terra, R., Díaz, A., & Zorrilla, J. (2010). Considering the information of the Niño 3.4 index in the operation
of the electrical system of Uruguay. Proceedings of the 33rd IAEE International Conference, Río de Janeiro.
Chaer, R. (2011). Fundamentos de modelo CEGH de procesos estocásticos multivariables. IIE-Reporte Técnico
SimSEE.
Kim, H., Vitart, F., & Waliser, D. E. (2018). Prediction of the Madden–Julian oscillation: A review. Journal of Climate,
31(23), 9425-9443.
Maciel, F., Terra, R., & Chaer, R. (2015). Economic impact of considering El Niño‐Southern Oscillation on the
representation of streamow in an electric system simulator. International Journal of Climatology, 35(14), 4094-
4102.
Ungerovich, M., Barreiro, M., & Masoller, C. (2021). Inuence of Madden–Julian Oscillation on extreme rainfall
events in spring in southern Uruguay. International Journal of Climatology, 41(5), 3339-3351.
Wheeler, M. C., & Hendon, H. H. (2004). An all-season real-time multivariate MJO index: Development of an index
for monitoring and prediction. Monthly Weather Review, 132(8), 1917-1932.
Zhang, C. (2005). Madden‐Julian oscillation. Reviews of Geophysics, 43(2).
Sección otros artículos
