Water is key to life on Earth. The distribution and quantity of precipitation controls the availability of water, yet little is known about past changes in precipitation. This is especially true at regional scales. The land region of the Northern Hemisphere, especially the mid-latitude region, offers an ideal opportunity to compare and contrast output from climate models with our longest and most comprehensive precipitation observations. This thesis develops current understanding, with the aid of climate models, to attribute changes in global mean precipitation to known key forcings. Perhaps the most obvious feature of twentieth-century global mean precipitation change is a decrease in response to mid-twentieth-century aerosol forcing. Changes in historical precipitation for the land mean of two regions in the Northern Hemisphere are shown to strongly resemble changes in the global mean, due to a greater sensitivity to aerosol forcing than greenhouse gas forcing. This aerosol response is predictable across models, which offer an ideal resource to test this constraint due to their large range in aerosol forcing. However, this aerosol response is not evident in these key mid-latitude precipitation observations. Observed runoff changes, derived from river discharge measurements, also contrast with observed precipitation changes in this region. This contrast is a consequence of an obvious breakpoint in the runoff-precipitation relationship. An ensemble of land surface models that are driven with observed precipitation data fails to simulate this contrast and breakpoint. Combined, these two lines of evidence strongly suggest that Northern Hemisphere mid-latitude precipitation observations are unreliable, at least in the early twentieth century. It is expected that the true trend is disguised by inhomogeneities. This should be recognised in future research that is reliant on these data.