 So to kick things off in our discussion of ecological forecasting, I want to start with the basic question, why forecast? Why should ecologists care about forecasting ecological processes and in particular, why might we be interested in making near term projections? So I'm going to focus a lot over these lectures on making forecasts that are both near term and iterative. In near term, I mean over timescales from subdaily to daily, weekly, monthly, seasonal, inter-annual, up to decadal, largely referring to timescales that are relevant to decision making and timescales where we can actually validate or verify how well we are doing at making predictions. One thing that's really important is that when we gain new information about how the world works, we want to incorporate that new information to forecast, so envisioning an idea where we make projections and predictions, we then gain new observations and we want the ability to incorporate those new observations and new information into forecasts. So one of the main reasons that we are interested in ecological forecasting is that there's a lot of environmental decisions being made every day in society from individual citizens up through industries and agencies that are not always being made with the best available data and best available information about how ecological systems work and how they may respond to different decision alternatives. Fundamentally, decision making is about the future, about how we think things will happen in response to different decision choices or if we make no decision. Because decisions are fundamentally about the future, often the best information that we can bring to the decision making process are explicit predictions and projections of how we think things will change in the future, which are fundamentally our forecasts. That said, I don't think that ecological forecasting is solely about decision making. I think ecological forecasting is fundamentally a win-win that has the potential to both improve environmental decision making, make ecology and environmental sciences more relevant to society, while at the same time accelerating and improving our basic science. So one of the reasons I think ecological forecasting can improve basic science is that I believe this forecast cycle, this idea of iteratively improving our forecasts as new information becomes available is in many ways a very strong expression of how we think the basic scientific methods work. When the scientific method we start with a hypothesis, which from a forecasting perspective is the models we use to make forecasts, they are models or an embodiment of our current understanding and current hypotheses and theories about how systems work. When we make projections with models, those are essentially the predictions that we make from our hypotheses that we then want to test against the world. When we confront those forecasts with new observations, we are testing our hypotheses, and then when we bring that new information into the models to improve our forecasts, we are updating our hypotheses and updating our understanding of the world. In many traditional ecological studies, this cycle of making hypotheses and testing them against the world occurs on the time scale of a research grant, so I might get a grant for two to five years where the research will often go out and collect data for most of that period and then analyze the data at the end of the project. By contrast, in near-term ecological forecasting, the idea is that we are continually confronting our hypotheses with observations from reality rather than saving them all up to the end. So this is the potential we believe to really accelerate the pace at which we do basic science by iterating through this process of hypothesis testing much more frequently and much more often. By analogy, if we look at the field of numerical weather forecasting, I believe one of the reasons that that field has gotten considerably better over time is they are essentially making a global projection of how they think the atmosphere works every six hours. Because of that, they're able to confront their hypotheses about how the world works four times a day, and they've been doing this for 60 years. And with that kind of feedback, they've gained a lot of understanding of how the system works. Another benefit of ecological forecasting for basic science is that forecasts require that you make projections and predictions explicit before new observations are made. So one of the challenges that faces a lot of scientific fields right now is something called the crisis of reproducibility, that we have a lot of results across all sciences, ecology is not exempt from this, where we can fit models and test hypotheses that often are challenged by new data. And one of the reasons that that happens is because it's very easy to overfit statistical models when all you're doing is looking at data from the past, because you can keep trying different models so you find the one that fits really well. But in many cases, you may be overfitting those models. When you make predictions into the future, you're less susceptible to that overfitting because you can't continually retune the model to things that haven't happened yet, to observations you don't have yet. You have to rely on more robust projections. Another advantage of ecological forecasting is it really forces us to do synthesis because it forces us to take the information we have and pull it together, assimilate it, and use that to make specific numerical projections, it kind of in many ways forces us to make our hypotheses much more specific and much more quantitative. So if you have very vague hypotheses and qualitative hypotheses, you can have a hypothesis that if I were to add nitrogen to a field of crops that the plants will grow faster. Well, that's a very general hypothesis because it doesn't say are the crops going to grow 5% faster or 25% faster or 50% faster. By contrast, if I make a specific numerical projection, then I can actually distinguish whether the observations I see in any particular experiment are consistent with what I understood about how the system worked from previous scientific work rather than just relying on a fairly vague null hypothesis of just that plants do not respond to nitrogen. So our traditional statistical testing just leaves us with a very, I think, unsatisfying test of does this have an effect or not, not necessarily is what we're seeing consistent with what we understood previously. Finally, one of the key parts of getting better ecology and improving our science is something that we have in common, I think, with all scientific disciplines that have attempted to make forecasts and attempted to mature to the point that they have some predictive understanding of the systems that they're working in, which is gaining that understanding requires feedback. So if you make predictions, but you don't learn how well you are at making those predictions, you don't really get better. So I think ecology right now is a lot of what we've been focusing on as a field has been projections over longer time scales, for example, predicting how the carbon cycle or specific species or specific systems might respond to climate change out to, say, 2100. So first, that is important and that's useful. But one of the challenges about that is that we don't actually learn very quickly whether we're good at predicting out to 2100 because we've never gotten any data from that period to see if we do it any well. By contrast, by making frequent short-term predictions, we can actually learn a lot about our ability to make predictions to get this feedback on how we are doing and use that to improve.