Hair Of The Dog: Technology for Climate Modelling

You can’t rely on reading tea-leaves for solving the present and upcoming concerns of climate change. With all the data centres, factories, e-waste and supply chains that technology weighs down the planet with, it only makes sense to use the problem, itself, as a solution. But it would take more than a hair-cut here.

Take a guess on this. Who is the world’s best meteorologist? Is s/he sitting in some super-computing facility? Or hovering above us in some satellite? Well, turns out this expert is busy singing on some tree in a busy forest. What’s Birdsong to us can be a complex and data-rich algorithm altogether in the world of these tiny creatures that are fluttering and chirruping their own codes. Take the Veeries from the forests of Northern US and Southern Canada. They are the best in the industry when it comes to predicting the Atlantic Hurricane season – thanks to their breeding patterns. Veeries have been observed to stop breeding early when the storms are earlier than expected. You watch them and you can get warnings – early enough.

In other parts of the world, in the Urban concrete jungles, computers of all sizes and stripes are trying to match up to sparrows and butterflies – in predicting hurricanes and other such events where fast action matters. Armed with swift attention and early alerts, humans can be better equipped to handle many Climate incidents.

Climate models and technology-backed intuition are trying to do just that.

It Takes Guts

The most interesting and beautiful use of our home-grown PARAM supercomputer could not have been any better – in this vein. It has been busy in taking a ‘guess’ but driven with supercomputing gut-feel. Anuman, which has been jointly developed by CES and VC & BA groups of C-DAC Pune, has been created to give hour-to-hour weather forecast over 50000 locations all over India using high-resolution weather model output – and it’s all generated using C-DAC’s National PARAM Supercomputer.

There is also the Saptarang initiative wherein CES Group (over last 20 years) has generated a range of simulated model outputs of meteorology, oceanography and air quality. The team has used a very high-resolution global forecast model that can be applied for seasonal forecast of Indian Summer monsoon rainfall along with global coupled model simulations for climate change studies. Also notable is the Experimental Extended Range Monsoon Prediction Experiment of the Department of Science and Technology, wherein the CES Group has helped with seasonal forecasts to India Meteorological Department (IMD) since 2005 using NCEP, Global T170L42 Spectral Model. CES, the Indian Institute of Tropical Meteorology (IITM), Pune and the National Physical Laboratory (NPL), New Delhi developed a methodology for conversion of large-scale emission inventories to gridded/ model-ready format over Indian region using GIS as a tool.

Climate models have seen improvements in accuracy, ability to handle complexity, and speed in crunching huge volumes of data- with technology advancements.

IIT Madras has also established a Centre for Atmospheric and Climate Sciences (CACS) dedicated towards studying the Earth system science and climate in Anthropocene. The idea is to understand the complex dynamics of the atmosphere and work towards strong fundamental and applied research to better understand and mitigate the impact of climate change on society. For example, one research studied Climate-change-induced risk mapping of the Indian Himalayan districts. It observed that the western Himalayan region is at more risk than the eastern Himalayan region in India. Another study explored how reduction in human activity can enhance the urban heat island - as insights from the COVID-19 lockdown. It was seen that the delay in winter crop harvesting during the lockdown increased surface vegetation cover, causing almost half the regional cooling via evapotranspiration. Turns out that since this cooling was higher for rural areas, the daytime surface urban heat island (SUHI) intensity increased (by 0.20–0.41 K) during a period of reduced human activity.

Step out of India, and you can find many such giant set-ups down in the weeds with a lot of simulations and forecasts whirring silently. Speaking of supercomputers, the first Cray petascale capabilities have been applied for forecasting and research – the most recent being selection by the European Centre for Medium Range Weather Forecasts (ECMWF) and Deutscher Wetterdienst (DWD) – for increased resolution and model enhancements.

There is also the Met Office Hadley Centre’s HadGEM3 family of models in the UK. And the NOAA Geophysical Fluid Dynamics Laboratory’s GFDL ESM2M Earth system model. Along with the Community Earth System Model (CESM) from National Center for Atmospheric Research (NCAR) in the US. Plus, exa-scale supercomputer Frontier which recently became available for scientific use and has helped to sharpen the resolution of Department of Energy (DOE)’s global climate model. It enabled it to simulate the fine-scale atmospheric processes that give rise to clouds (a big uncertainty in climate prediction work) - something that is left to guesswork in coarser models.

In UK, Met Office Cray XC40 supercomputing system and similar work are lauded for enhancing high-resolution climate modelling and to better assess future impacts of a changing climate at a regional scale, especially from high-impact weather and risk-resilience.

And of course, now there is AI in the fray too. Recently, we heard how NASA and IBM Research have developed a new AI model ‘Prithvi Weather Climate Foundational Model’ to be trained on a broad set of data (NASA data from NASA’s Modern Era Retrospective Analysis for Research and Applications or MERRA-2). It would be using AI to apply patterns gleaned from the initial data across a broad range of additional scenarios.

Guts need Pro-biotics

It all began with how sailors would predict storms with the patterns of winds they captured. When spread over a long period of time, any data captured with rigour and accuracy, can be a good telescope to see what may be expected ahead. Climate models are not new. But their accuracy, ability to handle complexity and above all, their speed and ease at crunching huge volumes of data- have been improved a lot due to some recent technology advancements. Specially those related to supercomputers, quantum computing and AI. This comes in handy when the goal is about simulating an entire ocean, continent or even the planet. The smallest improvement in spatial resolution (grid cell design and layout) of a model can call for a lot of fuel-tanks of compute power. If weather forecasts help us with hourly indicators, climate models play a game of many decades. Just think of the colossal number of calculations and crunching that all that prediction-work can take.

Imagine Earth as a giant, interconnected machine, its temperature, winds, and rainfall all intricately linked. Climate change throws a wrench into this machine, and understanding its future behavior becomes crucial. This is where climate modeling steps in, acting as our virtual time machine, offering glimpses of what’s to come explains Saurabh Rai, CEO, Arahas. “But unlike a crystal ball, climate modeling relies on hard science and cutting-edge technology. Early climate models were like trying to understand a complex recipe with key ingredients missing. They captured the broad strokes – the Earth heats up, oceans rise – but lacked the detail to predict regional variations. This is where parameterization comes in, the scientific art of translating complex, unseen processes (like cloud formation or ocean currents) into mathematical equations. Think of it as filling in the missing ingredients of the recipe. Today’s models are like having a seasoned chef in the kitchen. They incorporate real-world data from satellites and weather stations, along with sophisticated equations, to paint a more accurate picture of how Earth’s climate works.

Dhirender Mishra, Associate Vice President, Growth Advisory, Aranca avers that Climate modeling has become an essential tool for understanding and forecasting the impacts of climate change. “Significant advancements in climate modeling over the past few years have enabled a better understanding of the Earth’s climate system. These models have deepened our knowledge of climate dynamics and the potential future impacts of climate change. They are used to forecast future climate scenarios, which help in developing measures and policies to mitigate the potential impacts of climate change.”

Another key advancement is spatial resolution, Rai adds. “Think of Earth as a giant grid, with each square representing a specific location. Early models used a sparse grid, like looking at the world through a pixelated screen. This meant regional variations in climate were lost in the blur. Now, thanks to increased computing power, we can use much finer grids. Imagine zooming in on that pixelated screen, revealing the intricate details of a landscape. This increased resolution empowers communities directly impacted by climate change to make informed decisions about their future. Farmers can plan for changing rainfall patterns, and coastal communities can prepare for rising sea levels.”

Rai commends how India is at the forefront of this fight against climate change. “Recognizing the criticality of understanding regional climate impacts, they’ve developed their climate model, specifically designed to predict the impact on the crucial Indian monsoon – a lifeline for their agricultural sector.”

A BCT Digital-Chartis Research ESG and Climate Risk Survey shows that 72 per cent of global financial institutions plan to spend upto $500,000 or more on ESG technology.

And Pre-Biotics too

These models have their limitations and tight-ropes too.

A persistent challenge with many climate models has been the dilemma between accuracy and processing time, between simulation quality and speed – specially when a model tries to ‘leapfrog’  into the future.

Mishra reminds that there remain serious shortcomings in its predictions due to gaps in our fundamental understanding of the Earth system and the scientific limitations of supercomputing power. “However, with ongoing technological advancements, it is expected that these challenges will be mitigated in the future.”

Striking a balance between model complexity

and computational power is a delicate dance, underlines Rai. “Imagine trying to run a marathon on a treadmill – the more features you add (higher resolution, more complex parameterizations), the faster the treadmill needs to go (increased computing power). This can be a hurdle for developing countries with limited resources.”

It is critical to consider the problem of model-spread and multiple model outputs when making climate-change decisions.

He also talks about the challenge of model spread. “Imagine asking a group of meteorologists to predict the weather tomorrow – you might get a mix of sunny, cloudy, and rainy. Climate models are similar. While improved models with more sophisticated equations should theoretically converge on a clearer picture of the future, the inherent complexity of the climate system throws a wrench into that. Small differences in initial conditions can lead to a range of potential futures, creating a spread of possible outcomes. This highlights the need for continuous improvement and the importance of considering multiple model outputs when making climate change decisions.”

Then, there are practical walls and hoops to jump through too. As the BCT Digital-Chartis Research pointed out - over half of global financial institutions feel keeping up with changing regulations as the biggest ESG-related challenge. About 48 percent of the respondents pick here-risk assessment and mapping relevant ESG. There were also 48 percent who viewed integrating ESG into operational and financial workflows as a significant challenge. The key challenges that were distilled were as follows: meeting regulatory stress testing expectations (67 percent), accurate GHG (Greenhouse gas) accounting (56 percent), and integrating climate risk operationally into product lines (50 per cent).

Nonetheless, we need these climate models with more effort and capabilities.

As the optimistic Rai argues, Quantum computing could be the key to overcoming the computational bottleneck, allowing us to create incredibly detailed models with even finer resolution in a fraction of the time. “Additionally, new algorithms specifically designed for quantum computers could unlock even deeper insights into climate system dynamics. We could be looking at a future where understanding the intricate dance of clouds, oceans, and temperature becomes a reality, leading to more accurate and actionable climate predictions.”

Rai is right. “Climate modeling isn’t just about predicting the weather – it’s about shaping a more sustainable future for generations to come. It’s about giving humanity the tools it needs to navigate the complexities of a changing climate and build a more resilient planet.”

Technology can, for all its guilt in adding to its carbon footprint, aspire to be the world’s second-best meteorologist. One swallow never made a summer. 

By Pratima H

pratimah@cybermedia.co.in