The solar panels are removed before replacing the roof, then reinstalled afterward. Obviously, this costs extra, and the panels are not generating while they are uninstalled. But many (most?) roofing companies in places where solar is common also do solar, so they should be familiar with solar.
If you are considering solar, it is generally better to have it installed on a relatively new roof than to install it on an old roof that is likely to need replacement soon.
You call the installer. They de-energize the system and dismantle it, the roofer does their job, and then the panels get installed back.
Better have a newer roof! We replaced ours just prior to solar installation. The roof really had to be replaced.
The panels need to be removed first. Our company says it’s about $3000 to get that done. BUT they inspected the roof first, wanted to see our roofing contract, and won’t install panels on a roof that is old.
With my listed thermal coefficient. I’d need an average temperature of >150F higher in than summer than the spring/fall to have a 25% loss in average efficiency, which seems highly unlikely ,regardless of surface temperature vs air temperature.
However, I was curious, so I looked up some actual numbers of how production varied by temperature. I live in a moderate climate, without temperature extremes, but in early Sept last year, my area had a heat wave when average highs increased from 70s to 90s. Highs remained in the 90s for several days. The overwhelming majority of these days were sunny without clouds/rain. On the 90s days, peak generation dropped by roughly 4% compared to 70s days . In October, my area had a day with a clear, sunny day with a high near 100F. This time, generation dropped by ~3% on the 100F day over days with highs in the 60sF.
These types of 3-4% differences in output with temperature are much smaller than the variation from month to month due to other factors. For example my July average solar generation was ~70% higher than my October average generation, in spite of October averaging what I expect to be more optimal temperatures for peak efficiency.
Your Southern California/San Diego climate is ideal for solar power generation. That said, a panel surface can easily get heated up to 150 degrees Fahrenheit on a sunny, 100 degree day in places like Arizona.
See:
Throw in a less efficient panel, and there you go.
A 150F temperature doesn’t mean 25% loss of efficiency because the optimal efficiency temperature isn’t 0F. In any case, I didn’t say the panel can’t get to 150F. I said that the average temperature wouldn’t be 150F higher in summer than spring/fall.
I can read just fine. 
I am not taking about you. Did you see “throw in a less efficient panel”? 
The optimal efficiency is measured at 25 degrees Celsius. About 77 F. Not all panels have such low temperature coefficient as yours or mine. Some don’t need that - in places where it is generally cooler. When those are installed in Arizona though…
Well…I live in New England. Our air temps aren’t nearly 100 degrees, and it seldom gets below 0. We live where there is a good amount of sun most of the time. We shall see how this all goes!
I listed a specific temperature of 150F in my post, and you listed the same specific temperature of 150F in your reply to my post, so I assumed you were referencing it and referencing other content in my post. If you instead generally meant it’s possible for panels to reach 150F on a hot day in Arizona, and there is some unspecified degree of reduced efficiency that is larger for some panels than others, I don’t dispute any of that, nor did any comments in my posts.
Some will no doubt observe larger efficiency losses than the 3-4% loss I observed on days when air temperature highs were near 100F. The point is that for typical household solar installs, I expect these types of temperature-dependent efficiency losses below optimal will pale in comparison to losses from the other listed factors, as a source of variation in average daily/monthly solar generation.
A bit too much back and forth here. Please move on.
We have a flat rubber roof and we are adding a new layer soon. The panels are not bolted in- they are on individual metal supports with cement blocks on the supports- all movable to replace the roof.
IIRC my DS said it took him 7 yrs to recoup
This completely depends on the cost of electric in your region and your usage.
@cbreeze just asked about the Bay area
Even in any given area, that time to “profitability” can vary depending on many factors, like sun exposure, available roof surface (panels can’t be installed too close to roof edge depending on the roof pitch etc.), type of panels, etc.
Like with any home improvement project, get at least 3 bids from reputable contractors and compare their analyses. The online estimators can be way off. The contractors have access to much better tools than what is given to the public. In our case, the estimates are spot on so far.
and the state subsidy.
When I installed solar (about 13 years ago), each NJ SREC (which I generate every month and can sell in the SREC market) could be sold for $600 each. The market price dropped to as low as $100, before it recovered (it trades at around $215 or so in recent years).
Had I used PSEG financing at the time (which I didn’t), the program would have guaranteed the minimum price of $450 for each SREC. I obviously should have done that in retrospect. 
All good points. Was simply responding to @cbreeze’s request for someone to weigh in who had solars installed in the bay area 
There are a many factors, even within the same region of Bay Area… A key one that has not had much discussion is cost of the solar install itself, which is highly variable. In CA (including Bay Area), you can look up how much other people in your location paid for similar solar installs, using the NEM Application Database at CaliforniaDGStats . This includes filtering for a particular installer, showing the cost that installer charged for similar installs in your area, which I found to be helpful in price negotiation. In my area, prices are all over the map, rather than everyone paying about the same for similar systems. Lease vs own is also important for payback period. If you can afford it, owning often has far better net return than leasing.
For a variety of reasons, solar installers tend to estimate high on the amount of recommended panels. And if one does estimate high, this can notably increase payback time since excess solar generation over a full year period is paid at low wholesale electric rates, rather than high retail rates, under CA NEM 2.0.
For the system itself, the specific location of the panels on different houses can lead to large variation in payback period, on different houses in the same region. For example, if you have trees with shade and/or droppings, that can dramatically worsen payback period. Even things like which side of roof the panels are on is important. Most experts recommend south, but when I worked out the numbers, I had better ROI with my mostly west facing roof side (~250 degrees azimuth = WSW) than mostly south, due to benefiting from TOU electric plan, with higher electric rates in late afternoon when sun is setting and facing west. A battery can allow you to better tale advantaged of variable TOU rate differences, but tends to dramatically increase payout time overall since the TOU benefits are not enough to counter the often high cost of battery. So in general, having a battery notably lengthens payback period.
I mentioned PVWatts earlier, which is a free tool developed by the government National Renewable Energy Laboratory. PVWatts is good for rough estimations at the level of consumers. For more a far more detailed analysis, the NREL also makes the free tool System Advisor Model (https://sam.nrel.gov/ ), which had professional usage in mind and is used by government agencies, such as DOE. It allows things like listing the specific panel and inverter model and considering their technical performance parameters, 3D shading model, depreciation, etc. You can also load the info you entered from PVWatts, run simulations, etc. The simulations produce a variety of graphs including estimated net cash flow by year for 25 years, and how cash flow varies when you modify system or cost parameters. It also returns at what year expected payback occurs, and how that changes as you vary system, or if curious how payback year changes if you change location.
Yes, in the northern hemisphere, south facing maximizes total generation, but west or southwest facing maximizes the value of generation because electric demand is higher in the evening (which is why time-of-use pricing is higher in the evening).