Why Can’t We Better Prepare for Extreme Weather? | Catherine Nakalembe | TED

We can predict droughts, floods weeks, even months in advance, yet we still see the same crises unfold. Crop failure, economic and environmental devastation and displacement: the same crises that have trapped famine communities for generations. This is obviously not a prediction problem, it's a translation problem, one that I came to realize painfully in 2015. Equipped with the best tools available at the time, including that very expensive fancy drone, I spent August 2015 with my team in Karamoja, documenting yet another failed cropping season, one that I predicted months earlier, using satellite data. This was part of the worst drought in East Africa in decades, affecting 30 million people in Uganda, Kenya, Somalia and Ethiopia. After my field work, I did something researchers rarely do. I went straight to the office of the prime minister, and 24 hours after my second presentation to several ministers, food trucks were dispatched to Kamagaya on September 26th, 2015, exactly 10 years this week, which marked the first time the office used satellite data to trigger an emergency response. Following this, I helped design a program that would proactively release financing to support alternative employment for communities affected by drought. This program went on to support 450,000 people over five years, saving the government millions in emergency response and deploying several projects that included environmental restoration. But what haunted me then, and is still true today, is this. If we could mobilize emergency response within 24 hours, why couldn't we prevent this predictable crisis from unfolding? This paradox has deepened because today's capabilities make 2015's best look primitive. We have over 8,000 satellites and AI models and computation power that will make predictions, using this data with other data sets, to produce information at unprecedented scales and at unprecedented speeds. Yet, if you can combine this with advances in crop science... mobile banking, mechanization, the possibilities seem limitless. Yet just last year, in 2024, nearly one in three people were worried about where their next meal will come from. Climate disasters have more than doubled since the 1980s. So the question is: Why does this keep happening? I would like to tell you a story that will help you bridge the gap between why we have such incredible capabilities and are unable to deliver clear information for a farmer, for example to increase their yield, save their produce by reducing post-harvest losses, and having alternative income so they can survive through tough times. We have incredible technology, but we're missing translators to connect our predictions of that drought, for example, to real, tangible solutions that can get a farmer what they actually need to thrive. I'd like to share the story of Mary, whose experience represents millions of smallholder farmers around the world. Mary is not her real name, but she's a farmer in Iringa, Tanzania. But her story could easily be from Uganda, Madagascar or Senegal, or any other country where smallholders face similar challenges. Today's reality is this. For Mary and her neighbors, they plant February, March, for June, July harvests. This year, Mary acquired improved seeds, along with fertilizer that she heard about from a radio program. Unfortunately, despite her best hopes, rainfall was irregular, and she only harvested 800 kilograms from her one-acre plot. Her poultry business that used to provide critical backup income recently collapsed, so she does not have any savings. And it's just another year of surviving. Now imagine that Mary did in fact receive seasonal information sometime in January. Not only did it include the drought prediction, it included when and where she could access fertilizer, a recommended planting date, but most critical is that she has access to financing, that she could acquire a water pump so she could irrigate during dry spells. Come July, Mary has 3,000 kilograms. She has enough to see her through the next harvest

enough income, because she has access to buyers who provide premium prices for her produce, and storage so she can store it until markets stabilize. She can send her daughter to school, but most critically, she has extra income, so she can revive her poultry business. This is not science fiction. All the tools, all the technologies to get her that extra income exist today. So why is Mary still stuck? Why does she get set back by very predictable crises? The challenge lies in this messy middle, the complex web of relationships and real-life challenges that stand between our incredibly capable predictions and assessments and real, tangible solutions for Mary on the ground. For Mary, it's as if all our technology disappears into a black hole. And in my experience, drought predictions do not deliver pumps to the ground. They produce bulletins. Add to this complexity the fact that Mary has a small, irregular sized field that doesn't fit our perfect pixels. We are doing a terrible job mapping fields like Mary's. In addition to this, the basic infrastructure required for us to improve our predictions and really bring them to the ground are largely missing for regions like where Mary is based. This complex, messy middle is where all the capabilities shrivel, because it requires things that technology alone cannot provide. For example, it would require partnering with an extension agent who not only delivers fertilizer, trains a farmer and is an excellent data collector, but not replacing them. It would also require presenting our information in a way that is accessible to a bank, so that they can invest in a farmer like Mary, who needs to plant next month. So what is the path forward? We can either expand this messy middle, this translation gap, by creating more tech-driven silos, or we can use our current capabilities and venture to connect them to real solutions on the ground. To do this, there are five fundamental shifts that we would need to do. The first is we need to focus on translating, and this would require we're emphasizing reliability over perfection. A model that is 80 percent accurate, that delivers a pump to Mary, is far better than one that's 90 percent accurate, that never leaves our research paper or dashboard. It would require that not only do we fill that critical data gap, so we are better able to predict and assess the conditions in Mary's field, we would need to make sure our predictions can actually be evaluated. Number three, it would require shifting how we finance climate response, focusing on predictions that will get proactive responses so that Mary is able to recover her investment. Policies that encourage proactive planning are better than policies that emphasize emergency response. This would also mean we incentivize how we can connect our policymakers and people on the ground with the real advances our tools and technologies are able to provide. The fifth is people. We need to see people on the ground as accelerators, as people who are able to connect the real information that we're providing with real solutions -- improved seeds, irrigation infrastructure, etc. And I said five, but I have one more, and it's the most important. It's how we evaluate impact. Our combined effort cannot be measured by the number of projects or model accuracies. It should be measured by that extra income that helps Mary and uplifts her to become a resilient household. The technology to feed the world exists. Now we need to bridge this translation gap and move from data to decision, and prediction to prevention. Thank you. (Cheers and applause)