The North American Atmospheric Observing System (NAOS) Program is charged with providing the "basis for governmental decision processes on how to meet the evolving needs for atmospheric observations over... North America in support of the prediction and assessment of weather and of associated climate services". To fulfill its mandate, NAOS is developing scenarios for a future mix of measurement platforms under the so-called E1 hypothesis that: "It will be possible to reduce the number of radiosondes in the U.S. network without noticeably reducing forecast accuracy provided the sites removed have substitute observing systems in place". This hypothesis has been tested by the weather forecasting community with the aid of observing system simulation experiments. Here, we examine the consequences of the E1 hypothesis for climate change detection, by assuming that the 14 radiosonde stations targeted by NAOS for replacement by other observations are effectively removed from the climate monitoring network.

We estimate the effect of this potential observing system change on the multi-decadal linear trend in seasonal-mean 500-hPa temperature T integrated over the continental United States (CONUS) and over North America (NA). To do so, we take advantage of the relatively homogeneous NCEP-NCAR gridded reanalysis fields by sampling them over both CONUS and NA using three strategies: the full set of grid points, those closest to all radiosonde stations (SONDE), and those closest to these stations minus the 14 NAOS E1 sites (E1).

Trends in T during December 1958-November 1996 from each sampling strategy and in both regions are positive and seem reasonable compared to estimates directly from radiosondes. Only trends during the spring season (MAM) for some of the strategies are statistically significant. More importantly, differences in the trend of T among the sampling strategies are small, generally less than 0.1 K/decade. Examination of the trend field suggests that it is indeed of large enough scale to permit it to be well sampled by the current radiosonde network or one that is modestly smaller. We extend this result to a more extreme reduction of the existing network, namely the Global Climate Observing System (GCOS) Baseline Upper-Air Network, which includes only 6 (14) stations in CONUS (NA). Differences in trends for NA or CONUS between GCOS and SONDE are typically less than 15%, despite the large difference in station coverage.

We conclude that continental-scale estimates of trends in 500-hPa temperature based on the existing North American radiosonde network, or on a range of potential changes thereto, are robust. This result depends on the large scale of the underlying trend field considered here; other quantities of interest to the climate research and monitoring communities may be considerably more sensitive to the number and distribution of upper-air stations.