The monsoon season: May 2025 outlooks for India

Written by Gerard Castro, Tyler Anderson & Yazmina Zurita, part from Nebbo technical team.

This blog post is a continuation of our series, so far composed by the March and April outlooks. Here, we will look at the expected wind forecast for the months of May and June in the same 3 regions in India, as well as the “observations” (ERA5 reanalysis) for the past April.

Disclaimer: The rationale behind these outlooks is to present the capabilities and limitations of the most well-known seasonal models. These are obtained from C3S and bundled in a multi-model fashion, without any other complex postprocessing. In particular, a more indepth analysis would require features as those present in our site forecast product:

• Model downscaling to remove bias & variance issues, as well as reliability problems, by increasing the resolution
  • If no finer observational reference is used (no downscaling), the models’ resolution is 1º.
• Full probabilistic overview
• Re-forecast skill analysis & accounting for the forecast error

With that in mind, both the April ERA5 map as well as the India forecasts initialized this May 2025 are presented below. The latter forecasts correspond to our multi model aggregation and lead months 1 & 2: i.e. valid for the end of this May & June. 

For more information on the process followed in order to generate (& interpret) all the visualizations, one can refer to the corresponding section in our previous post.

Observations for April 2025: ERA5 reanalysis

Let us compare, in Figure 1, the forecasts for April 2025 that were made both in March (lead month 2) and April (lead month 1) against the observational data, which will come from ERA5.

Importantly, it should be noted that the forecasts are forecasting wind speeds at the 10 meter level, while the observations are the wind speeds at the 100 meter level. Although the wind speeds at these two levels are correlated and the anomalies should remove most of the differences, the fact they are not exactly the same also reminds the importance of a proper bias correction. 

Furthermore, the same region labels used in the previous posts will be adopted henceforth too. As a short reminder, the three regions are listed below:

• Region 1: the region composed of north Karnataka, south east Maharashtra and west Telangana.
• Region 2: the region composed of north Gujarat.
• Region 3: the region composed of east Kerala and West Tamil Nadu.

Fig 1. ERA5 100m wind speed anomaly map for India. In red, the regions to focus on. Baseline period: 1993-2016.

Let us perform an analysis on how Figure 1 compares to our previous forecasts segmenting it by region:

• Region 1:
  • In April 2025, Region 1 experienced atypically high wind speeds predominantly in the eastern sector, while the northwestern section registered slightly reduced wind velocities.
  • Analyzing the March forecasts, the ensemble mean (ensmean) anticipated a modest inclination towards increased wind activity throughout most of the area. Although it revealed a mild bias towards windier conditions, particularly in the central region, the observational data confirmed significant anomalies in the east with a positive anomaly also identified at the center.
  • April forecasts, however, predicted a broader and more uncertain situation; with slight negative wind speed anomaly across the region, but with almost equally populated terciles as the lower, normal & upper probabilities were almost equal. Except for a low wind speed anomaly realized in the southeast, the ensmean failed to align closely with actual observations, lacking significant signal differentiation.
  • In conclusion, March ensemble forecasts were more reflective of observed conditions, indicating captured signals in the western and central zones, whereas April forecasts displayed less confidence in any of the 3 outcomes (lower/normal/higher wind speeds) but with a less congruent ensmean.
• Region 2:
  • April 2025 observations for Region 2 revealed elevated wind speeds prevalent across the area, notably in the south.
  • Forecasts issued in March anticipated lower wind anomalies; however, probability plots exhibited weak signals for this prediction. Subsequent April forecasts maintained the projection of reduced wind activity, though probability mappings demonstrated insufficient signal across the majority of the region.
  • Ultimately, actual conditions showed heightened wind activity, particularly in the south, challenging the ensemble mean’s earlier predictions.
• Region 3:
  • April 2025 observations depicted low wind speed anomalies throughout Region 3, with notably diminished wind speeds central to the region.
  • March forecasts projected low wind speeds in the north and high speeds in the south, partially supported by probability maps in the southern sector.
  • Later forecasts in April adjusted predictions towards low wind speed anomalies ubiquitously, except for anticipated high wind incidences along the southern coast. Here, probability plots strongly indicated northern low wind anomalies, but signals for southern high wind activity were less discernible showing lower confidence than in the northern region.
  • In essence, forecasts successfully evolved by locating the higher anomaly in the south while approaching the forecast date, even though they inadequately represented its final exact location in the south. How high anomalies were slightly misplaced by the models can be clearly seen in Figure 2, which features a significantly high anomaly episode for April in the southern Indian shores and the closest side of the Indic ocean.

Fig 2. ERA5 100m wind speed anomaly for south of India. Baseline period: 1993-2016.

Summary

In short, the analysis of April 2025 wind speed anomalies across the three regions shows the forecasts were generally successful and captured sufficient signals, despite some discrepancies. 

The results also highlight the benefits of advanced model downscaling, as employed by Nebbo in its site forecasts. Such techniques could enhance the precision of forecasts, allowing for better capture of smaller signals and anomaly intensities. Incorporating these refined methods would elevate the utility of forecast data, supporting more informed decision-making and ensuring alignment with complex localized weather patterns.

Forecast for May 2025: lead month 1

To communicate the upcoming wind forecasts for India clearly, we start by looking at the ensemble mean anomaly map in Figure 3. This map shows areas with higher wind speeds in red and lower wind speeds in blue.

Fig 3. Multi-model wind speed anomaly (ensmean). In red, the regions to focus on. Baseline period: 1993-2016.

Although it highlights expected changes in wind speed across India, it doesn’t indicate how likely these changes are. For that, we turn to Figure 4, which breaks down the probability of experiencing less windy, normal, or windier conditions.

Fig 4. Multi-model probability. From left to right: low, normal & high wind probabilities. Baseline period: 1993-2016.

Now, let us focus on the three regions. Apart from a specific analysis, below the forecasts initialized in May are compared with those initialized in March & initialized in April which are also valid by the end of May:

• Region 1 
  • The forecast indicates a robust signal for normal wind conditions, with windier-than-normal scenarios being largely dismissed across much of the region.
  • Comparing this with previous predictions (lead months 3 and 2), there has been a notable shift towards normal wind speeds from initially higher predictions.

Fig 5. Multi-model probabilities for region 1 (from left to right: probability of having less, average & more wind).

• Region 2 
  • The current forecast suggests a stronger likelihood of windier conditions in the western part of this region, though the rest of the area remains uncertain, with probabilities evenly distributed across different outcomes.
  • In lead month 3, the forecast was leaning towards windier conditions. In contrast, lead month 2 showed predictions that are very similar to the current lead month 1 forecast.

Fig 6.  Multi-model probabilities for region 2 (from left to right: probability of having less, average & more wind).

• Region 3 
  • In this region, normal wind conditions are highly likely, with windier scenarios being mostly discarded.
  • The confidence in both less windy and normal conditions has progressively increased from lead month 3 to 1, contrary to the decreasing probability for windier-than-normal conditions.

Fig 7. Multi-model probabilities for region 3 (from left to right: probability of having less, average & more wind).

Forecast for June 2025: lead month 2

We continue our analysis by focusing on the forecast for June with lead month 2. By applying a similar process, we derive the ensemble mean anomaly for India, as illustrated in Figure 8, accompanied by the probability maps in Figure 9.

Fig 8. Multi-model wind speed anomaly (ensmean). In red, the regions to focus on. Baseline period: 1993-2016.

Fig 9. Multi-model probability. From left to right: low, normal & high wind probabilities. Baseline period: 1993-2016.

Now, let us focus on the three regions. Apart from a specific analysis, below the forecasts initialized in May are compared with those initialized in April and also valid by the end of June:

• Region 1 
  • The northern area is likely to experience reduced wind speeds, with the possibility of windier conditions largely excluded, while the southern part is expected to maintain normal wind conditions.
  • Compared to lead month 3, the probability of experiencing less windy and normal conditions has risen, while the likelihood of windier conditions has diminished.

Fig 10. Multi-model probabilities for region 1 (from left to right: probability of having less, average & more wind).

• Region 2 
  • Across the entire region, there is a strong signal for less windy conditions, effectively ruling out the occurrence of windier conditions.
  • In comparison to the lead month 3 forecast, the expectation for less windy conditions has strengthened, whereas the probability of windier conditions has significantly decreased.

Fig 11.  Multi-model probabilities for region 2 (from left to right: probability of having less, average & more wind).

• Region 3:
  • It is likely that the majority of the region will experience less windy conditions; however, there remains a notable likelihood of normal or windier conditions occurring as well.
  • In contrast to the lead month 3 forecast, the probability of less windy conditions has increased, while the likelihood of windier conditions has decreased.

Fig 12. Multi-model probabilities for region 3 (from left to right: probability of having less, average & more wind).

Summary