Here, the authors describe the construction of a forcing dataset for Land Surface Models (including both physical and biogeochemical models; LSMs) with eight meteorological variables for the 35 year period from 1979 to 2013. The dataset is intended for use in a model intercomparison (MIP) study, called GTMIP, which is a part of the Japanese-funded Arctic Climate Change research project. In order to prepare a set of site-fitted forcing data for LSMs with realistic yet continuous entries (i.e. without missing data), four observational sites across the pan-Arctic region (Fairbanks, Tiksi, Yakutsk, and Kevo) were selected to construct a blended dataset using both global reanalysis and observational data. Marked improvements were found in the diurnal cycles of surface air temperature and humidity, wind speed, and precipitation. The datasets and participation in GTMIP are open to the scientific community (https://ads.nipr.ac.jp/gtmip/gtmip.html). This data includes four data sets below. *L0 data: 34-year, reanalysis-based (ERA-Interim), 30-min data *L1 data: L0 data merged with the GRENE-TEA site observations *Detrended data: 20-year detrended driving data for spin up based on L1 (1980-1999) *FPAR/LAI data: MODIS based monthly data at GRENE-TEA site (2001-2011; monthly)