Stream Temperatures Along the River Continuum - AI Edition
- Jason G. Freund
- 5 days ago
- 7 min read
Updated: 4 days ago
AI and large language models are everywhere, for better or worse. I am keeping an open mind but know that there are some serious issues. First I think we need to acknowledge the the genie is out of the bottle and AI is going to almost certainly become a larger part of our lives in the future. Again, for better or worse. The expansion of AI to our daily lives has been "sneaky". It is how your cell phone predicts the next word(s) you are likely to type. It is embedded into nearly every program you use now. And it is used by my students way more than it should be or at least in ways it should not be used (of course, they're not the only ones using AI poorly and not assessing its results with a critical eye).
I recently wrote a deeply-researched post - it's kind of my thing - about how stream temperatures change along the river continuum. It was a 7-minute read, according to Wix, the software I use to create this website/blog. I used ChatGPT to write a similar outline and then turn that outline into a post once I was happy with were the outline was going with the post. As an aside, one of the major issues with AI use is that people (i.e. my students) use it to save themselves time but they often do not put in the time and effort to get a good product from AI. In my experiences with AI, like most things in life, you get out of it what you put into it. Creating a good product is the result of iterative prompts where the user modifies and adds to their prompts which allows AI to provide better results. Using AI "lazily" provides a poor result much of the time. One of the images I'll use below spelled a number of words incorrectly - but my next couple of prompts fixed that (mostly).
As you read the post below, you will see it is not as detailed as I was - it is about a 3-minute read. Maybe you see that as an advantage of AI - it is concise. However, a common failing of AI is that, without much additional prompting, its answers are often a mile wide and an inch deep. To get it to use citations, I had to add additional prompts. To its credit, ChatGPT is now using real sources and not hallucinating them (though I still plug my student's sources into Google Scholar to be sure they are real - and that's not always true).
Below is ChatGPT's attempt at essentially the same post I created. This came with some coercion from me for it to dig deeper than its initial attempts which were pretty poor. With yet more effort, I could have gotten it to dig deeper into the topic.
Beyond the Continuum: How Springs, Shade, and Tributaries Shape Stream Temperatures
Introduction
Stream temperatures are often thought to follow a simple downstream gradient: cold in the headwaters, gradually warming as the river widens and flows toward its mouth. This idea is rooted in the River Continuum Concept (RCC), which describes predictable patterns in stream structure and biology from source to sea (Vannote et al. 1980). According to the RCC, temperature, light, and nutrient inputs increase predictably downstream, helping to structure the aquatic communities found along the way. But when it comes to stream temperature, reality is messier—and much more interesting. Springs, tributaries, riparian shading, and geomorphic features interrupt this predictable pattern, creating a patchwork of warm and cold zones that change across space and time. These thermal complexities have major implications for stream ecology, especially for coldwater species like trout.
The Expected Pattern: A Temperature Gradient from Source to Mouth
Under ideal conditions—constant flow, unbroken canopy, and no significant groundwater or human disturbance—stream temperatures increase as the river moves downstream. This is largely due to increasing solar radiation, wider channels that reduce shading, and greater air-water heat exchange (Poole & Berman 2001).
Springs, Shade, and Tributaries: Breaking the Continuum
Real-world streams rarely follow this neat progression. Instead, local features frequently override longitudinal temperature trends.
Springs and Groundwater Inputs
Springs act as natural air conditioners. Groundwater temperature remains relatively constant year-round, providing a powerful buffering effect on stream temperatures (Constantz 1998). Spring-fed reaches stay cooler than surrounding segments during summer, creating thermal refuges for coldwater fish (Ebersole et al. 2003).
Tributary Inputs
Tributaries influence mainstem temperatures in complex ways. A spring-fed tributary can cool downstream segments, while urban or agricultural tributaries may raise temperatures. These create mixing zones and ecological boundaries (Rice et al. 2001; Fullerton et al. 2015).
Riparian Zones and Canopy Cover
Shade reduces solar radiation. Dense riparian forests limit daytime warming (Moore et al. 2005). Canopy loss leads to thermal increases (Johnson 2004).
Channel Morphology and Flow
Shallow, wide channels warm faster than deep, narrow ones due to higher surface-area-to-volume ratio (Poole & Berman 2001; Dugdale et al. 2015).
Human Impacts
Dams and impoundments retain heat, warming downstream waters (Lessard & Hayes 2003). Urban and agricultural runoff alters stream temperature. Restoration can help (Krause et al. 2005).
Seasonal Context: When Timing Matters
In summer, shaded and spring-fed reaches are critical refuges. Snowmelt and leaf-out affect thermal dynamics (Acuña et al. 2008; Caissie 2006).
Why It Matters: Ecology and Conservation
Coldwater species depend on thermal refugia. As climate change raises temperatures, maintaining access to cold tributaries and springs is essential (Isaak et al. 2016; Ebersole et al. 2001).
Conclusion
The RCC remains a valuable framework, but stream temperature is more than a linear pattern. Springs, shade, and tributaries create a thermal mosaic vital for stream life.
References
Acuña, V., et al. (2008). Hydrological Processes, 22(16), 2185–2194.
Beechie, T., et al. (2013). BioScience, 60(3), 209–222.
Caissie, D. (2006). Freshwater Biology, 51(8), 1389–1406.
Constantz, J. (1998). Water Resources Research, 34(7), 1609–1615.
Dugdale, S. J., et al. (2015). Hydrological Processes, 29(23), 4453–4460.
Ebersole, J. L., et al. (2001). Transactions of the American Fisheries Society, 130(4), 733–746
Ebersole, J. L., et al. (2003). Canadian Journal of Fisheries and Aquatic Sciences, 60(7), 901–910.
Fullerton, A. H., et al. (2015). Freshwater Science, 34(1), 23–35.
Isaak, D. J., et al. (2016). Fisheries, 41(7), 374–386.
Johnson, S. L. (2004). Canadian Journal of Fisheries and Aquatic Sciences, 61(6), 913–923.
Krause, S., et al. (2005). Hydrological Processes, 19(3), 741–756.
Lessard, J. L., & Hayes, D. B. (2003). River Research and Applications, 19(7), 721–732.
Moore, R. D., et al. (2005). Journal of the American Water Resources Association, 41(4), 813–834.
Nelson, K. C., & Palmer, M. A. (2007). Journal of the American Water Resources Association, 43(2), 440–452.
Poole, G. C., & Berman, C. H. (2001). Environmental Management, 27(6), 787–802.
Rice, S. P., et al. (2001). Canadian Journal of Fisheries and Aquatic Sciences, 58(3), 518–526.
Stewart, J. S., et al. (2021). Ecological Applications, 31(3), e02254.
Torgersen, C. E., et al. (2012). Fisheries, 37(10), 477–491.
Vannote, R. L., et al. (1980). Canadian Journal of Fisheries and Aquatic Sciences, 37(1), 130–137.
Evaluating and Using AI
The above AI written post has the information correct, but it is in no way a fun read. It provides no context, just facts, and that is rather boring. I know, I read a lot of scientific papers which are rarely fun to read. I also read a lot of student writing. However, we recently replaced a lab writing project because it was evident that in the age of AI, it was not always achieving our goals of having students actually think and write their own papers. Instead, we created pencil-to-paper, in lab assignments - no computers or phones allowed - that seem to have thwarted poorly "written" AI papers. What is old, is new again.
I am not anti-AI, but at the same time, I am not a fan of things that reduce thinking and "the productive struggle". We learn by doing things, making mistakes, and understanding those mistakes so we do not make them again. We tired of seeing students use AI as the easy way out. More importantly, we tired of seeing students use AI without thinking. They get an answer and assume it is correct. AI is designed to give an answer, not a correct answer. In fact that is a failing of AI, it is unable to say, "I don't know", three very important words. If there is one thing I have experienced with AI, it is that creating a good product is an iterative process that requires you to think about how to write better prompts to get a desired result. And it requires you to evaluate the product that it returns. Some were not learning because they used AI to replace the efforts that go with the struggle. Like any tool, using AI correctly, not lazily, is required but too often it replaces thinking and the struggle.
While I know many before have failed to predict the future, I think it is pretty easy to see that AI is going to be a huge part of our future. Can we use AI in ways that helps create knowledge and our understanding of the world around us? It is capable of finding patterns that our human brains can not. It has revolutionized parts of science and society already. At the same time, AI does not think critically about what it produces. And if we fail to think critically about what it produces, it becomes an unreliable tool.
AI Will Not Replace Me...
But at the same time, I plan to use AI more in the future. But not to fret, I do not plan to use it to write posts - other than a post here and there about AI like this one. I find AI very interesting and I am curious to see where it goes. Again, for better or worse. Most importantly, I will never pass AI off as my own writing. However, AI can help organize thoughts, create figures to go with posts, and who knows what else I will find it works for. One of the places I have found AI to be a huge benefit is in finding and summarizing peer reviewed papers on topics I am interested in. I have included those files below and I am increasingly using them as a starting point to help me better understand a topic so I can dig deeper into that topic.
There is no question that AI is getting better and more useful all the time. There is also no question that AI is still rather imperfect. We'll see what the future holds.
Links to AI-Generated Annotated Bibliographies
Expect some of these things to be future topics - others are things that I am searching because they are research topics I am interested in. I find one of the most useful things that AI has done for me is help find articles that are more difficult to find in our library databases and Google Scholar. Though, I will say, I found and read most of these articles before AI found them. It did do a pretty good job of summarizing the articles.