Compelling evidence from direct measurements of thermospheric density and ionospheric temperature indicates that the terrestrial upper atmosphere is experiencing long-term cooling over the last few solar cycles. Such cooling seems consistent qualitatively with speculation of upper atmospheric changes associated with anthropogenic increase of greenhouse gases as first suggested by Roble and Dickson (1989). However, quantitative differences among observations and between simulations and observations of the long-term cooling still exist, raising important questions regarding the most significant driver(s) of climate change at ionosphere and thermosphere altitudes. In this presentation, we will review ionospheric long-term trend observations based on incoherent scatter radars. In particular, we will present new observations from two high latitude sites at Sondrestrom (Invariant latitude 73.2°N) from 1990-2015, and Chatanika/Poker Flat (Invariant latitude 65.9°N) over the span of 1976-2015 (with a gap from 1983-2006). Results are compared to conditions at the mid-latitude Millstone Hill site (Invariant latitude 52.8°N) from 1968-2015. The aggregate radar observations have very comparable and consistent altitude dependence of long-term cooling trends. The trends below 275 km are significantly higher (-3 ~ -1K/year at 250km) than anticipated from model predictions given the anthropogenic increase of greenhouse gases. The cooling trends above 275 km continue to increase in magnitude but values are strongly dependent on magnetic latitude. Our results indicate that drivers, from above of below, other than those directly associated with greenhouse gas increases can play an important role in determining the strong long-term cooling in the upper atmosphere.