Learning time series analysis in hydrogeology Good morning,
I know there are some similar questions, but I'm asking anyway since I didn't find a suitable answer for what I'm looking for. I've got some daily hydrological data about a spring, spanning for 5 years, and I'd like to analyse them in order to 


*

*Find cause-effect relations between input series (aka rainfall, snow, external sources) and output series (mainly discharge, but also temperature and conductivity)

*Check for long-term memory effects (autocorrelation)


I'm unfortunately not a statistics-savvy guy, I'm an earth science student with some average knowledge of calculus, statistics and programming (web languages like PHP, JS, also some Python and C), but I need to perform these analyses and report them in a medium-to-short time frame (a couple of months).
I've already done the basic time plots, auto-correlation, cross-correlation and power spectral density charts for each series and they show some things, but seasonality (recurring cycles at 6 and 12 months) mostly belittles all else for now.
What are the best techniques for analyzing the data? What path would you follow? How and where can I educate myself enough to understand what has to be done? Please give me honest even if harsh feedback guys, thank you.
 A: The cause-effect theory is a broad field with many approaches developed so far, as an example it started from Granger causality which was later modified to accommodate non-stationary data. Then there is Shanon entropy based information flow. The Liang-Kleeman information flow theory and application, wavelet coherence and transform and so many... in the list. Many atmospheric studies had been conducted with the above methods and published in SCI-Indexed journals which can be accessed with the above keywords. As far as correlation and autocorrelation is concerned, the above methods has own steps before applications but kindly keep in mind that causality is correlation but correlation is not causality. So making any statement based on these methods should be align with assumptions and limitations of each method since you are using climate data which is quite non-stationary and chaotic. The signal in both station and gridded data is varying but the signal is there, which obviously is the GUY we usually look for. These suggestions are my own opinion based on my  self study, you can make your choice after reading few of these methods. 
