I have a time-series record of groundwater levels missing c.a. 20% of the data and I want to know if are there suitable methods for assessing monotonic trend.
You can use non-parametric Mann-Kendal test that can analyze time series data with missing values. Moreover, it is one of the good method to detecting trends in hydrological data like groundwater level, rainfall etc. You may use software R to perform the analysis.
You can use interpolation functions using softwares such as R or MATLAB. These functions are useful for finding cells of a matrix with missing data (NaN), interpolating the missing value through the first value before and after the missing value. Kriging tool in GIS is based on the mentioned functions.
Dear Dr. Towe, Thanks for your answer. I am sure this kind of question has been asked on RG given the fact that this is an "easy level one". But I couldn't find any with the specificities that I need to understand.
Dear Rahman, thanks for your answer. I am aware of the usefulness of np Mann-Kendal trend test for hydrogeological time-series. I was just wondering if there would be any limits for missing values and/or long gaps in data for the usage of this method. Perhaps another methodology would be more suitable in my case. I'm just wondering. I am using a very user-friendly freeware software called Past (please, see the link below).
Dear Ana, thanks for your answer. I believe that interpolation won't work in my case because I have an actual huge gap in the middle of my time-series, instead of missing data cells. Expertise-based intuition tells me that the data should follow a linear trend. But I can't really determine that, since apparently, the groundwater levels don't follow a linear response to precipitation events in my case.
You may also consider co-kriging if you have any variable somehow related with your time series. In this case, regardless of limit of missing data, it will work out perfectly.