ModE-RA is a global monthly paleo-reanalysis of the modern era from 1421 to 2008, using an offline data assimilation approach to reconstruct past climate fields by blending model simulations and observations. In the early period, natural proxies and documentary data are used, while from the 17th century onward, instrumental measurements are also assimilated. The impact of each observation is stored in an observation feedback archive, providing information on preprocessing steps and forward models. ModE-RA includes estimates of key climate fields and a reconstruction, ModE-RAclim, to disentangle the roles of observations and model forcings. ModE-RA is suitable for studying intra-annual to multi-decadal climate variability and past extreme events. The study uses an ensemble-based Kalman filter approach, combining the newest data compilations with an ensemble of atmospheric general circulation model simulations. Previous reconstructions date back to 1601, but this study uses more data, improved data treatment, and a new method for assimilating short time series. From 1421 to 1657, the observational network is based on proxy records and documentary evidence, while from 1658 onward, instrumental measurements become available. The assimilated observations include temperature-sensitive and precipitation-sensitive proxy records, documentary evidence, and early instrumental measurements. Other climate fields such as geopotential height and wind components are also reconstructed. The study evaluates ModE-RA against independent time series, gridded instrumental observations, and other climate reconstructions. It shows good agreement with 20th-century gridded monthly products for temperature, sea-level pressure, and precipitation. ModE-RAclim is used to assess the effect of observations and model forcings. The study also provides an observation feedback archive to trace the effect of observations and input data. The study uses an ensemble-based Kalman filter approach, combining model simulations and observations to produce a 600-year-long global monthly atmospheric paleo-reanalysis. The study uses the ensemble square root filter (EnSRF) approach, which combines the observation information with the model simulations. The background-error covariance matrix is calculated from the ensemble members, and spatial localization is used to limit the distribution of observational information. The study uses atmospheric model simulations based on the ECHAM6 model, with different transient monthly-varying forcings and boundary conditions. The study also uses observational data from proxy records, documentary data, and instrumental measurements. The study uses a variety of proxy records, including tree rings, coral, ice cores, lake sediments, speleothems, and bivalves. The study also uses documentary data, including indices and phenological data. The study uses instrumental measurements on land and from ships, including pressure measurements from ships. The study uses a variety of data sources, including proxy records, documentary data, and instrumental measurements. The study uses aModE-RA is a global monthly paleo-reanalysis of the modern era from 1421 to 2008, using an offline data assimilation approach to reconstruct past climate fields by blending model simulations and observations. In the early period, natural proxies and documentary data are used, while from the 17th century onward, instrumental measurements are also assimilated. The impact of each observation is stored in an observation feedback archive, providing information on preprocessing steps and forward models. ModE-RA includes estimates of key climate fields and a reconstruction, ModE-RAclim, to disentangle the roles of observations and model forcings. ModE-RA is suitable for studying intra-annual to multi-decadal climate variability and past extreme events. The study uses an ensemble-based Kalman filter approach, combining the newest data compilations with an ensemble of atmospheric general circulation model simulations. Previous reconstructions date back to 1601, but this study uses more data, improved data treatment, and a new method for assimilating short time series. From 1421 to 1657, the observational network is based on proxy records and documentary evidence, while from 1658 onward, instrumental measurements become available. The assimilated observations include temperature-sensitive and precipitation-sensitive proxy records, documentary evidence, and early instrumental measurements. Other climate fields such as geopotential height and wind components are also reconstructed. The study evaluates ModE-RA against independent time series, gridded instrumental observations, and other climate reconstructions. It shows good agreement with 20th-century gridded monthly products for temperature, sea-level pressure, and precipitation. ModE-RAclim is used to assess the effect of observations and model forcings. The study also provides an observation feedback archive to trace the effect of observations and input data. The study uses an ensemble-based Kalman filter approach, combining model simulations and observations to produce a 600-year-long global monthly atmospheric paleo-reanalysis. The study uses the ensemble square root filter (EnSRF) approach, which combines the observation information with the model simulations. The background-error covariance matrix is calculated from the ensemble members, and spatial localization is used to limit the distribution of observational information. The study uses atmospheric model simulations based on the ECHAM6 model, with different transient monthly-varying forcings and boundary conditions. The study also uses observational data from proxy records, documentary data, and instrumental measurements. The study uses a variety of proxy records, including tree rings, coral, ice cores, lake sediments, speleothems, and bivalves. The study also uses documentary data, including indices and phenological data. The study uses instrumental measurements on land and from ships, including pressure measurements from ships. The study uses a variety of data sources, including proxy records, documentary data, and instrumental measurements. The study uses a