Relationships between SST indices and all India rainfall can be quantified through multiple regression analysis. We perform a multiple regression analysis of all India rainfall against four SST indices: Nino 3.4, the Indian Ocean dipole index, SSTs in the tropical Atlantic and the trans-Nino index. Here, we compare the observed regressions to the regression of the ensemble mean indices.
The teleconnection to ENSO is well represented, with an observed regression of -0.79 and an GloSea5-GC2 regression of -0.69.
The relationship with the Indian Ocean dipole is too weak in the model, with an observed regression of 1.15 and an GloSea5-GC2 regression of 0.32.
The teleconnection to the tropical Atlantic is appears not to be represented in the model, with an observed regression of -0.62 and an GloSea5-GC2 regression of 0.42 that has an especially large standard error of 0.37.
2. Climatological monsoon biases
Predicting South Asian monsoon precipitation and circulation on time scales of
weeks to the season ahead remains a challenge. Current state-of-the-art GCMs
contain large biases, particularly dryness over India, which evolve rapidly from
initialization and persist into centennial length climate integrations. We present
initial results from our Ministry of Earth Sciences Indian Monsoon Mission
collaboration project to assess and improve weekly-to-seasonal forecasts in the
Met Office Unified Model coupled initialized Global Seasonal Prediction System
(GloSea5-GC2, [1]). Using a 20-year, 15-24 ensemble member hindcast set in
which atmosphere, ocean, sea-ice and soil moisture are initialized from May start
dates, we assess the monsoon seasonal prediction skill and the mechanisms
contributing to skill in GloSea5-GC2.
The Met Office seasonal forecast system GloSea5-GC2 predicts
monsoon rainfall over India well in some years, but also has
prominent forecast busts.
The largest driver of Indian monsoon seasonal variability is ENSO.
GloSea5-GC2 predicts ENSO anomalies, and their teleconnection
to India, well. Other factors, like the Indian Ocean dipole and the
tropical Atlantic, are represented less well.
Future work will use bias correction experiments to determine
influence of mean state errors on prediction skill.
Summary
1. Background and GloSea5-GC2
References (1) Williams et al. 2015, Geosci. Model. Dev. Discuss, 8, 521-565 (2) Rajeevan et al. 2012, Climate Dynamics, 38, 2257-2274
3. JJA precipitation prediction skill
4. Influence of ENSO & IOD
3. JJA time-series
5. Future work
JJA precipitation JJA SST
GloSea5-GC2 shows rainfall deficits over India and the Maritime Continent and excess rainfall over the western equatorial Indian Ocean and the Western North Pacific.
Equatorial cold tongue SST biases are present in all basins. In the Indian Ocean this presents as a positive Indian Ocean dipole bias.
ENSEMBLES MMM and CMAP
JJAS precipitation correlation map
Correlation maps show significant (p > 0.05) skill over the Maritime Continent and equatorial Pacific, consistent with past studies [2].
Correlation of JJA ensemble mean all India rainfall with GPCP: 0.41
GloSea5 and GPCP JJA
precipitation correlation map
We will use the results of our assessment to motivate bias correction
experiments, such as wind stress correction or nudging experiments, in different
basins in GloSea5-GC2. We will use this framework to test whether the existing
equatorial cold SST biases are causing the poor relationships with the Indian
Ocean and tropical Atlantic SST variability. Through these experiments we will
determine the impact of mean state biases on forecast skill, and test the utility of
bias correction techniques for operational implementation.
JJA all India rainfall anomaly time-series
JJA Nino 3.4 SST anomaly time-series
ENSO is the most important driver of Indian monsoon rainfall interannual variability.
ENSO anomalies well predicted by GloSea5-GC2.
Indian rainfall anomalies are less well predicted and have larger ensemble spread.
Relationships between JJA indices