The inhabitants of a small city carved into the side of a cliff are blind and use echolocation to navigate. There is a lot of empty space, around a kilometer of between the cliff side where the village resides and the other cliff side, in other words a 1km valley. Will echolocation still work effectively in this kind of situation?
Yes, in combination with other skills
Your residents are going to use a combination of techniques to navigate. Echolocation is a way of avoiding obstacles, not of navigating over longer distances. In the Wikipedia image below, the bat can tell that there's a box ahead on its left, but the bat has no idea what's on the other side of that box.
If your villagers want to walk from the church to the blacksmith shop, for example, they'll have to have an idea of where both places are located and generally how to get between them. Here's a great overview of how blind people in the real world navigate from Chicago Lighthouse:
When in familiar places, visually impaired people generally know the layout and memorize where things are. Learning to travel in different or unfamiliar places is done by using orientation and mobility (O&M) skills. Orientation is the actual planning of how to get to and from places. Blind and visually impaired people use other senses – like sound, touch and smell – to orient ourselves to our surroundings. So, if I am walking outside and know there is a school nearby, then I can assume that it is near when I hear children playing and laughing. Likewise, the smell of freshly baked bread and cookies tells me I am near a bakery.
By adding echolocation into the mix, you're giving your people a much better ability to avoid walking into objects. But they still won't be able to get around using that skill alone. They'll start hearing the hammering of the blacksmith and know that they're getting closer to the destination. As they approach, they'll be able to use echolocation to avoid running into the anvil.
There are some exceptional real worlds cases of humans using echolocation, take a look at https://en.wikipedia.org/wiki/Human_echolocation to see a few. There are videos where people avoid obstacles, sink basket balls and take hikes along rocky paths. In a demonstration in a documentary I can't recall the name of Ben Underwood could even successfully detect small objects on a table, and he could distinguish wooden objects from metal ones . Most of these people are exceptional, in the sense that they discovered their ability and worked it out themselves. In your community though, echolocation is normal. IRL exceptional skills are probably normal in your cave, and the exceptional cases would be really cool. Finding a hanging rope to climb somewhere? No problem! Workers throwing bags of turnips to each other from a cart? Sure! Spear hunting? Why not? Spear hunting fish? Maybe.
As has been noted though, the resolution of echolocation is courser than that of eyesight, and this becomes a problem over large distances; you could compare it to a low-resolution camera where far away objects are eventually within the same pixel. Also, over large distances sound is distorted, and reflections, wind etc will compromise perception. There will be some sound though, so people can probably "see" some blurry picture at least. If gravel, smooth granite, limestone, rough granite and slime-covered, moist granite all give of distinct echoes, you could maybe tell enough about far off structures to get a sense of direction, and if nothing else know what type of local environment you are in; though echolocation often assumes that the person using it makes the sounds themselves, the way in which surfaces reflect ambient sounds are important as well.
Also, like @CAE Jones and @Andrew point out, many places make distinct sounds by themselves. Playing children, dripping water or a smithy will all be little beacons which will help people navigate.
Let's also not forget the possibility of various hearing aids. Fairly simple parabolas can produce surprisingly focused beams of sound, and one could easily imagine these people using cones, tubes and various other shapes to direct, magnify and otherwise manipulate the sound according to their needs. This is not to speak of all the possible devices to generate special purpous sounds for certain types of echoes, or devices that detect certain sounds. All these aids, ranging from the equivalents of glasses or binoculars to observatories, would be either carried by people who need them, kept in most homes or stationed in certain buildings, maybe watchtowers.
This said, other senses would also be important. Except for touch, in the form of rails along paths where echoes are impractical and so on, I think foremost of heat. If you are outside on a sunny day, you can feel where the sun is and use that to navigate much longer distances, but even more miniscule differences could be useful. Does the cave become progressively colder the further in you get? Can you feel the difference in radiation from the warm opening of the cave from it's cold interior? There could also be differences in the behaviour and temperature of the wind depending on what your surroundings look like.
Anecdotally: probably. Compare the Horseshoe Canyon Ranch, which is in the Ozarks. The Louisiana Center for the Blind takes blind trainees there annually, with lots of room for said trainees to wander off unsupervised between (and sometimes during) scheduled activities.
More relevant to the question (it isn't precisely a city carved into a rocky canyon, after all), there was this time when I was there with a group of mixed skill levels, returning to the lodge from the rock-climbing area. In terms of layout, the lodge is in the middle of a shallower slope, at the bottom of a steeper section where visitor cabins are, and the climbing area is up a completely different, steep and rocky slope, across a gap which I assume is around 0.25km or so in width based on how long it took to cross. On this particular evening, we'd taken a different path for some reason, rather than the more direct gravel road connecting these areas, instead going down the rocks and crossing a big meadow . Things were going slowly, and some people were needing extra assistance (lots of big rocks and such on the way down), and for this I was quite useless, and permitted to go on ahead if I wanted. Somehow, absolutely nothing went wrong. It turns out that even Ozark-sized mountains close together are sufficient to navigate by, without requiring additional clicks (the ambient acoustics were quite sufficient), and as I got closer, voices from the lodge were audible. I also suspect that, had the lodge been silent and I'd veered away from it too far to catch echoes from the building, it was still sufficiently apparent which side of it I was on, so I'd've just followed the slope/road, whichever I'd've hit first.
There was also this small playground behind the lodge, not directly in contact with any sidewalk or other structure people would likely frequent. The equipment was not very tall, as playground equipment goes, and it was little more than one family-sized structure in the middle of a box of gravel. Knowing that it was behind the lodge somewhere, and that nobody else was paying attention to it was sufficient to find it, though I will point out that the information was necessary (it was not simple to pick out ambiance reflecting off it before getting very close to cane's reach of the conspicuous boundary).
While my travel/mountain skills might have been unusual at the time, my echolocation is on the poor side - useful, but not reliable. The most reliable day-to-day passive echolocation is hard to describe, just because it's always there in the background. The most salient example in a non-canyon context, for me, would probably be my front yard. I'm far enough from the road, with enough trees, that it's easy to fail to distinguish my house from trees or the neighbors' fence, especially if there's not enough sun/traffic and I'm away from the driveway. The clearest cue is the garbage can, because it has a very distinct echo profile, being just large enough, and having the right material/hollowness to be picked out from farther away than, say, a random person of the same height. The driveway also ends without sidewalk quite a way before reaching it, but normally I can echolocate it before reaching the end of the driveway. Which, of course, means that if I'm coming back from taking the garbage to the road for pickup, it's a lot harder to find the way back. (LCB helped a lot with this particular problem, and I eventually did have a stone path installed from the driveway to the porch, but before those things, the echoes from that thing were a rather big deal).
All this is with fairly passive echolocation (not that I didn't try clicks / snaps once or twice when looking for the playground / garbage can). People who train to echolocate learn how to use different clicks to different ends, for things like range vs detail / area, etc. My experience is something of a lower bound, and a full-blown city with generations of experience to build on would outperform the one impressive feat to my name on a daily basis. The biggest consideration I'd keep in mind is range and levels of detail, and range will depend on both the size of objects and whether or not there are other object around to obscure them.
One thing I have never encountered, even though it makes a ton of sense to me, is strategic use of windchimes. Neither the LCB apartment complex, nor any of the blind staff's houses had any such decorations whatsoever. No school for the blind with a multibuilding campus that I have ever visited has utilized chimes. The NFB national convention uses ushers who form a path of criers, shouting directions (rather than going the Disneyland route and having speakers replay the same message from the 70's on a loop), and that is the most I have ever encountered as regards deliberate audio signage. It's enough to make me question whether or not a canyon city full of blind people would bother, even though someone with, say, a shop in the middle of the canyon would probably want to draw attention to themselves somehow. (The NFB convention's market / exhibit hall just had a maze of long tables with Braille labels to indicate what you were passing. You'd have to know to constantly check the table for labels, of course.)
I expect the real answer is "yes, but do not rely exclusively on echolocation; it has gaping weaknesses, and temperature / noise / environment can influence it positively or negatively."
The power level of the echolocation signal in bats is similar to that of the sounds the human voice can produce (order of 100 decibels). This allows the animals to perceive objects over distances never recorded to be further than 20 meters. https://pubmed.ncbi.nlm.nih.gov/22978903/
Regardless bats have also specially adapted receiving organ devices. The better human echolocator I've ever withnessed was that on Stan Lee's show "Superhumans". A blind man who could slowly drive a bike in a silent parking lot.