In our study, we use MADA data alongside a hydrological model (to convert MADA dryness / wetness into standardised streamflow anomalies), which allows us to place recent observed droughts in the Ganga basin in a long-term historical context, showing how unusual recent decades have been compared to natural monsoon variability over the last 1,300 years. Importantly, MADA does not rely on modern instruments or models, but on natural climate archives themselves.
When we compare the 1991-2020 drying (using 30-year backward-moving means of standardised streamflow anomalies) against all other 30-year windows back to 700 CE, the recent period is the most negative in the record.
In that sense, the “worst in 1,300 years” refers to the relative position of 1991–2020 within the reconstructed distribution, not just a single-year extreme.
Regarding uncertainty: The reconstruction is built as an ensemble (multiple plausible reconstructions), and the paper reports a 95 per cent confidence interval envelope from that ensemble. The key point is that the recent multi-decadal drying sits outside the range of comparable events across the pre-instrumental portion of the record, even accounting for uncertainty.
That said, there are important caveats: The underlying proxy network (tree-ring data used to reconstruct drought metrics) is less dense earlier in time, especially prior to the 14th century, which can reduce fidelity for very early centuries.
AS: You have also highlighted that most climate models fail to capture this drying trend in the Ganga? Why is that and how can it be improved?
KT: A major result is that while state-of-the-art CMIP6 models generally reproduce the warming trend, most do not reproduce the observed drying / streamflow decline in that specific spatial domain over recent decades. In our analysis, only a small subset of models captures the drying tendency, and even those tend to underestimate the magnitude. Why might this happen? Here are some possible reasons:
South Asian monsoon rainfall is sensitive to anthropogenic aerosols (and their spatial distribution), which are still challenging to represent accurately in models.
Land–atmosphere processes and irrigation: Irrigation and land-use changes can alter surface fluxes and regional temperature gradients in ways that can affect monsoon circulation; these processes are often simplified or inconsistently represented.
Resolution and regional dynamics: The monsoon involves sharp gradients, orography, mesoscale convection, complex cloud physics, and air–sea coupling that coarse-resolution global models struggle with.
Improving these aspects in the model might help improve accurate spatial simulation.
AS: What does paleo-climatological research inform us about the major river systems of India, such as the Brahmaputra, Narmada, Godavari, Krishna and others? Is similar drying / droughts being observed there as well?
KT: Paleo-hydroclimate information exists for parts of India, but coverage is uneven by basin and by proxy type. More studies like ours are needed to look at how coarse-scale proxy field reconstructions can be coupled to local hydrological forcing.
Moreover, even though most subcontinental rainfall is controlled by the summer monsoon rains, the Indian subcontinent does not behave as one coherent hydroclimate unit (especially concerning rain outside northern hemisphere summer).