The International Mass Loading Service provides displacements due to four loadings (in decreasing their magnitude): 1) ocean tidal loading atmospheric pressure loading; 2) atmospheric pressure loading; 3) land water storage loading; 4) non-tidal ocean loading. You need to apply all of them.
The time series contain harmonic variations with periods more than 1.5 days, namely annual and semi-annual consitiuents, but they do not contain harmonic variations with periods shorter than 1.5 days.
Depends on the problem you want to solve. These loading variations are real, and their omission in data reduction will cause systematic errors. So, in general, both loading time series and loading harmonic variatinos should be applied
There are three ways:
First, fill the form with downloading pre-computed loading. Set the output type "wget control file". Submit it. Then examine the long URL in your browser. Copy this line. Then find the start and stop dates. Replace the start and stop dates and execute this url as an argument of wget. You need embrace the URL with apostrophes. Command wget will get you the list of wget commands. Execute this list. Then write a simple program (or script) in whichever language your prefer that combines these two steps.
All models are good. Models that failed validation are not put in the International Mass Loading Service. Read the papers with model description, ask your colleagues, toss a coin. Reanalysis MERRA2 models start at 1980.01.01, run through present ans has latency 15–50 days. They are supposed to be the most stable models, in a sense that no changes in the processing chain is made. Semi-operational GEOSFPIT models start at 2000.01.01, run through and have latency 6–15 hours. They are considered less stable than MERRA2.
The models that I use in routine data analysis: ATM_GEOSFPIT, LWS_GEOSFPIT, TOC_FES2014B, EQUIL01, NTO_OMCT. For epochs before 2000.01.01 when GEOSFPIT is unavailable, I use ATM_MERRA2 and LWS_MERRA2. I use harmonic variations at diurnal and sub-diurnal bands for all these loadings, except for the non-tidal loading. I use for data reduction only the S1 contribution to harmonics variations of non-tidal loading model OMCT05 (i.e. the variation with the period exactly one day: 86400 s).