Earlier I repeated the common comment that climate is always changing, with some of my own thoughts about what that may or may not mean. Time to think about science itself, and how it is always changing. Same as the fact that climate is always changing doesn't mean that we can necessarily ignore the current changes, we can't always count on science to change in a way that we might (for other reasons) like. On the extreme, there are folks who say 'since science is always changing, so you can't tell me that I'm wrong to think that the moon is made of green cheese.' Replace moon being made of green cheese with any of very man other statements that are equally incorrect, today and for the future. While the science changes, there are some fairly specific ways in which it changes that exclude a number of those hoped-for conclusions.
One part that changes little or not at all are the observations themselves. Once you've got the observations, there they are. If your backyard thermometer says 125 F at a certain time, then that's what it said -- as best the thermometer and clock could be read. But what does that reading mean? Well, that's subject to some change. We may discover that the type of thermometer you used is error-prone, or biased, or has a bias that grows with age, or that you put it in direct noontime sun, or that it's right over your grill or dryer vent, or ... So, while the reading itself won't be changing, trying to make use of it quite possibly will. The meaning of the observation may well change as we learn more. Quite possibly in time we'll discover that your thermometer's reading doesn't tell us anything we want to know about the atmosphere, as such, no matter how informative it turns out to be about when you have cook outs. That's one avenue of change in science -- observations of some kinds are discovered to be unreliable for what we're trying to do. Conversely, observations that we didn't realize were useful for some problems, someone bright discovers how to make them useful in that new area. It took over a decade, but the MSU is now useful for studying some things about climate.
A non-change is labeling. Recently, Pluto was reclassified as not being a planet after 70+ years of being called a planet. Absolutely nothing meaningful about Pluto changed -- its orbit takes the same time, its size is the same, its composition, estimated history, and so on, are all still the same. Calling it a planetoid or plutinoid or whatever makes no difference to the science as such. In biology, classification can be a much more serious matter, but I'll let biologists discuss that.
Some sorts of things change, and poor science teaching or journalism would have you think it was 'big'. But that's the weakness of the teaching or reporting. When I was much younger, Pluto had no moon, Jupiter had 12. Now, Pluto has 1 moon (Chiron) and it's huge compared to what it orbits, and Jupiter and Saturn now have so many 'moons' that there's another labelling argument about just what constitutes a 'moon'. In the bad teaching, students have to memorize trivia like how many moons Jupiter has. Tomorrow or next week, we'll discover another. But this doesn't change the science any. The science isn't in the list of moons. It is in things like 'why are there moons?', 'how does a body hang on to a moon?' , 'how long can a body hang on to a moon?'. (I sometimes teach astronomy. Numbering the moons never comes up on my tests; explaining why Jupiter has any does. The science of this won't be changing drastically, but the number of moons will. I want my students to focus on the more fundamental parts.)
A different sort of change, but not one that necessarily causes difficulties for prior science, is getting more and better observations. We've got vastly more meteorological data today thanks to satellites and radar systems than 40 years ago. Yet ... high and low pressure are still high and low pressure, storms happen, hurricanes happen, and we are still limited in how far ahead weather (as opposed to climate) can be forecast. What changed are things like how close to the limits we can come (and we do do a lot better than 40 years ago!). Sometimes we get enough new observations that a simple earlier explanation about, say, why there isn't much water vapor in the stratosphere turns out to be wrong. The fact that there isn't much water vapor up there doesn't change. But the explanation changed when we got enough new data to disprove the old one.
Where we see a lot of rapid change is in areas where we go from unable to say much -- because we have no suitable observations -- to being able to see the things. Astronomy has gained enormously from this in the last 40 years. Before, we were stuck with telescopes that were looking through the turbulent earthly atmosphere, that filtered out ultraviolet, most infrared, and all x-ray and gamma-ray energy. Now, we can see all those things from satellite and discover ... tons of things. (Yay!) But when looking at new things, though, we tend not to change much about what we used to think. We thought those things about things we could observe in the first place. Sometimes the new look, as with more data on stratospheric water vapor, does help us change our explanations. For biology, being able to sequence DNA has been another revolution in being able to see things. But we already knew that DNA existed and had a major role in heredity. What changed isn't that sort of thing, but being able to see details and to compare details. (Again, Yay!)
Three major revolutions of the 20th century for which I know something good about the science are quantum mechanics, relativity, and plate tectonics. By the way, three books to read are Thirty Years That Shook Physics by George Gamow, Relativity by Albert Einstein, and On the Origin of Continents and Oceans by Alfred Wegener (4th edition in English is available from Dover). All three are by prime players, and are readable by nonspecialists. In the case of Wegener, his ideas (Continental Drift) are not the revolution that actually happened (Plate Tectonics), but it's extremely interesting to read.
In all three cases, the theories, ideas, and methods that existed before the revolution were retained -- in the areas they'd been successfully tested in. When I try to examine how radiation is bent around a rain drop or scattered off a gas molecule, I use Maxwell's equations, not Quantum Electrodynamics (the part of quantum theory you'd use). For this problem, Maxwell's pre-quantum equations are sufficient. When we look at atmospheric circulation, we use Newton's equations, not Einstein's. If wind speeds were near the speed of light, we'd have to pay attention to Einstein and abandon Newton. Fortunately, wind speeds top out at 300 m/s, rather than 300,000,000 m/s, and we can do the simpler physics. Plate tectonics was a spectacular revolution, in the sense of unifying a number of observations (locations of mountains, volcanoes, and earthquakes, for instance) with a simple idea (plates bashing in to each other, floating around at the surface of the earth). But ... nothing about the age of the earth, the composition of the various plates, thhe history of the rocks, etc. was changed by this revolution.
Particularly in quantum mechanics and relativity, this is called 'convergence'. As conditions move from the extreme conditions that lead to the new idea, to the conditions the old ones were tested in, the new idea has to give the same (at least within old observational tolerances) answers as the old method. This gives some sharp limits on just how large a revolution you can count on for the future. For instance, the earth was known to be much older than 6000 years by the early 1800s. That isn't going to be changing -- there are too many different lines (whether quantum mechanics, classical mechanics, plate tectonics, mineralogy, ...) that point to a much older earth. Now maybe the current (if I remember correctly) 4.55 billion years will get revised to 4.6 or 4.5 billion. But a 6000 year old earth is gone for good from science. Ditto a flat earth, or getting rid of the greenhouse effect, or a number of other things.
Since things do change, however, we're in a position to be informed spectators about some of them. Some areas have unsettled questions, and we can watch the progress in getting them nailed down. One which has seen some recent progress is expectable sea level change. After observations that some glaciers were moving and melting much faster than previously expected, estimates of sea level change for the next century went from fairly well constrained (we thought) to 'well, something like 50 cm, not counting whatever the things we don't know very well any more might do.' That left the door open for possibly not much addition (if the things that surprised us were already done being surprising) to quite a lot of addition (if they could continue speeding up). The current estimate by Pfeffer and others is now for 0.8 to 2 meters (2.5 to 7 feet) of sea level rise by 2100. This will be changing as more people look at just what the (current) uncertainties are, and what factors might be coming in to play. Keep your eyes on this one. On the other hand, it's extraordinarily unlikely that the sea level would start dropping.