I noticed a mistake I made in the EEPROM lifespan calculation, each value is only measured every 24 hours. If I use something like an offset counter that would have to get refresh every hour but if I just clear all the values after sending them to the server I can reduce the writes to twice a day so 2/24h. 1m/ 2 are 500.000 days which are 1368 years. So that will work.
After the power down I unplugged the ESP8266 and all other components so the battery could charge via solar power. I also directly connected the solar panel and the battery since the power breaks down to ~5V if enough current flows. I did add a diode in between to ensure there is no backflow, the voltage drop is roughly 1V so the battery is not in danger (at least I hope so). Today the charging LED is no longer on, so the battery is full.
My last post was at the 26th (about the power down) this means it took 17 days to recharge. The Battery has 3000mAh at 5V. 17 days are 408 hours. 3000mAh / 408h = 7.3mA . We can double that at day to 14mA and at night we got 0mAh. This means we have to use less than 176mAh per day.
Another major problem is: the soil resistance is very inaccurate. I think one of the wires is broken or the electrodes (steel screws) have connection issues. If the sensor is never detection water, I will probably just drown my plant, so that needs fixing.
I have not digged to deeply into SPI which the ESP8266 yet but I did a lot of SPI projects with the STM32F4 and STM32F3 so I hope its not that hard.