Why are journals telling me that my article is plagiarized even though I wrote it by myself?

Chances are the journal ran your paper through an automated plagiarism checker (the only realistic way to check for plagiarism these days) and this 70% number is what the program found. That you found nothing using Turnitin isn't enough evidence to prove there was no plagiarism: a plagiarism checker is only as good as the data it has, and it's possible Turnitin doesn't have the original paper.

Having said that the journal should tell you which paper they think you've plagiarized from, as well as which sentences are plagiarized. If they don't, you should absolutely write back to ask. From your comments you've already done that, so there's nothing to do now except wait. The fact that at least two journals have checked your paper and found plagiarism is a bad sign; on the bright side, you can ask both journals for more details and it's less likely they both don't answer.

If they respond you'll be able to fix the plagiarism if it's there, or point out why there's no plagiarism if it isn't there. If they don't respond after a reasonable time, then the only thing left to do is submit the paper elsewhere.

Springer and Elsevier both use cross check (see Springer's webpage and Elsevier's webpage). In particular, the Springer link has a note under "tips and tricks - how to interpret the results" that states

iThenticate is a tool detecting text similarities. It does not detect plagiarism. The similarities may be an indication of plagiarism.

and

When you find a very high similarity percentage (for example 80%) for an article that has already been published it is most likely that the match found is to the paper itself. The source needs to be excluded from the matches. Unless it was published somewhere else than Springer.

It also notes that

In addition different internet sources including repositories and Wiki’s such as Wikipedia are indexed.

With that in mind, it should be enough to write the editor politely noting that you have made a pre-print available and it appears the similarity detector has found that rather than any actual plagiarism. Since you've taken those steps there's nothing to do but wait and politely follow up in a couple of weeks.

Since data science and text analysis has blown up recently ( this decade ), Plagiraism checkers have been a huge use for it.

Problem is, text analysis can range from the very rudimentary (check how many times a word appears in a document and come up with a ratio of it to the all the words in the document) to the very complex (look for sentence structures, topic analysis, etc).

You would think that journals would run some high-end plagiraism checker, but you never know. They may be running a rudimentary checker that just racks up all the words in your paper and it seems to pattern-match the ratio of words you use compared to your whole document to seem similar enough to another person's document. But, that wouldn't take into account the order in which the words are used.

I would ask them what analysis software they're using. Seeing as it's a journal, and science is all about reproducability, they should be able to direct you to the plagiraism checker to let you run it through yourself and see what it's doing and also provide in-depth knowledge of HOW it's doing it.

Because the biggest concern in data science (and all it's applications .. including text analysis, ai, machine learning) .. is you want to know how the inner workings of something work. That way you're not at the mercy of some "black box" spitting out some answer that everyone is taking at face value (when there could be HUGE amounts of bias or overly-simplified stupidity going on INSIDE the black box).

I would also worry about anyone you may have shared your work with.. even in confidence.

I worked with a group one time where we shared our work internally. But, one guy in the group was friends with some folks in another group. They were having issues with something, so they asked him for help. Like most folks, he was lazy, so instead of walking through everything with them.. he just showed them our work.

Later on professor was asking several groups to step forward to figure out who did original work and who plagiaraized. Because what this one group was doing was asking ALL the groups in class if they could "help" them with different parts.. then just stitching everyone's work together and submitting it as their own.

The guy in our group did not say he was showing our work to anyone. I operate in a group environment under the unspoken assumption that it's just common courtesy to not show a group's work to any outside parties unless you get the inner-group's permission first. So, I was floored that this guy was sharing our work w/o our knowledge.. and it had vast repercussions. (Because at the school I go to, the honor code states that the folks who got plagiarized are just as responsible for the folks that plagiarized.)

So, I would think about who you possibly showed your work to. You may have thought you showed it to them in confidence, but they could be wandering around showing others.

And if you did this work for a class.. a professor may have taken your work and turned it in as their own. That's not uncommon.

This semester a professor is letting another prof show up and toss a mini data science project onto us. I've had this other prof before, and he's the type that asks students to do in-depth research projects in his field of interest, and there were rumors of him just turning around and publishing their work as his own after he fleshed things out a bit more. Well, the project he's tossing on us is another one of these "I have no clue how to do this, so I task you with dreaming up a way.. oh, and document and detail everything as if it's a research paper." I find it very shady.. and assume he's just going to take everyone's work, condense it down, and publish it under his own name after-the-fact.

So... think about who may have had access to your work. There's some underhanded, shady folks out there.