Abstract The performance of the operational extended range forecast (ERF) issued during the 2023 monsoon season is evaluated using two modelling frameworks, viz., India Meteorological Department’s CFSv2-based ERF (hereafter called IMD ERF) and extended range prediction (ERP) based on unified model runs at the National Centre for Medium-Range Weather Forecast (hereafter called NCM ERP). The multi-model ensemble (MME) forecast is also prepared based on these two modelling systems to see if there is any improvement in forecast skill in the MME ERF. The real-time ERFs using three modelling systems (IMD ERF, NCM ERP, and MME ERF) have captured the observed intra-seasonal variability of monsoon 2023 very well with the lead time of 2–3 weeks. Comparing the two ERF systems for all India rainfall, it is observed that the IMD ERF performed slightly better in week-1 and week-2 forecasts, whereas the NCM ERP performed slightly better in week-3 and week-4 forecasts. Further, it is observed that the MME ERF performed slightly better in week-3 and week-4 forecasts compared to both modelling systems. With respect to the prediction of the active-break cycle of monsoon over central India, the IMD ERF performed better than the NCM ERP in forecasting the active and break phases of the monsoon over central India, with the CCs remaining significant up to three weeks. The MME ERF showed a slight improvement, particularly in the week-3 forecast, compared to both modelling systems (IMD ERF and NCM ERP). The verification of categorical forecasts at the met-subdivision level for four weeks using three categories (above normal, normal, and below normal) also shows slightly improved skill in MME ERF compared to IMD ERF and NCM ERP systems, particularly in week-3 and week-4 forecasts. Research highlights <UnorderedList Mark="Bullet"> <ItemContent> The real-time extended range forecasts from IMD ERF, NCM ERP, and MME ERF successfully captured the intra-seasonal variability of the 2023 monsoon over India with a lead time of 2–3 weeks. </ItemContent> <ItemContent> These operational ERF systems also captured the onset, active-break cycle and withdrawal phases of monsoon very well. </ItemContent> <ItemContent> IMD ERF performed better than NCM ERP in weeks 1–2, while NCM ERP showed better skill in weeks 3–4. The MME ERF offered limited improvement in weeks 1–2 but performed sligthtly better than individual models in weeks 3–4. </ItemContent> <ItemContent> Improving the skill of operational ERFs beyond two weeks at meteorological-subdivision level would greatly enhance their utility for planning and decision-making in sectors such as agriculture, reservoir water management, power etc. </ItemContent> </UnorderedList>

Multi-model ensemble-based extended range forecasts based on IMD’s and NCMRWF’s operational modelling systems

期刊论文
发表日期：

2026年2月9日

查看全文

D R Pattanaik,
C J Johny,
Hemadri Bhusan Amat,
Ashish Alone,
Ankur Gupta,
Akhilesh Mishra

摘要

Abstract

The performance of the operational extended range forecast (ERF) issued during the 2023 monsoon season is evaluated using two modelling frameworks, viz., India Meteorological Department’s CFSv2-based ERF (hereafter called IMD ERF) and extended range prediction (ERP) based on unified model runs at the National Centre for Medium-Range Weather Forecast (hereafter called NCM ERP). The multi-model ensemble (MME) forecast is also prepared based on these two modelling systems to see if there is any improvement in forecast skill in the MME ERF. The real-time ERFs using three modelling systems (IMD ERF, NCM ERP, and MME ERF) have captured the observed intra-seasonal variability of monsoon 2023 very well with the lead time of 2–3 weeks. Comparing the two ERF systems for all India rainfall, it is observed that the IMD ERF performed slightly better in week-1 and week-2 forecasts, whereas the NCM ERP performed slightly better in week-3 and week-4 forecasts. Further, it is observed that the MME ERF performed slightly better in week-3 and week-4 forecasts compared to both modelling systems. With respect to the prediction of the active-break cycle of monsoon over central India, the IMD ERF performed better than the NCM ERP in forecasting the active and break phases of the monsoon over central India, with the CCs remaining significant up to three weeks. The MME ERF showed a slight improvement, particularly in the week-3 forecast, compared to both modelling systems (IMD ERF and NCM ERP). The verification of categorical forecasts at the met-subdivision level for four weeks using three categories (above normal, normal, and below normal) also shows slightly improved skill in MME ERF compared to IMD ERF and NCM ERP systems, particularly in week-3 and week-4 forecasts.

Research highlights

The real-time extended range forecasts from IMD ERF, NCM ERP, and MME ERF successfully captured the intra-seasonal variability of the 2023 monsoon over India with a lead time of 2–3 weeks.

These operational ERF systems also captured the onset, active-break cycle and withdrawal phases of monsoon very well.

IMD ERF performed better than NCM ERP in weeks 1–2, while NCM ERP showed better skill in weeks 3–4. The MME ERF offered limited improvement in weeks 1–2 but performed sligthtly better than individual models in weeks 3–4.

Improving the skill of operational ERFs beyond two weeks at meteorological-subdivision level would greatly enhance their utility for planning and decision-making in sectors such as agriculture, reservoir water management, power etc.