Earth science professors

Three parts of your work where AI is already doing real lifting, and what stays yours.

• Field notebook1845–1900

  Field notebook, hand lens, and transit (geological survey era)

  The equipment of the 19th-century earth-science professor was the same as the working geologist or astronomer: a Brunton compass, hand lens, rock hammer, field notebook, and for the astronomer, a transit telescope with attached chronograph. Course content was delivered by lecture and demonstration; laboratories were collections of minerals, fossils, and astronomical charts. Fieldwork was central from the start -- professors like Florence Bascom led students into the field, and field mapping was taught as a graduate skill from the 1890s onward.

• Radiosonde1919–1950

  Radiosonde and upper-air sounding networks (synoptic meteorology era)

  The radiosonde -- a balloon-borne instrument package transmitting temperature, pressure, and humidity data -- was developed in Europe in the early 1930s and adopted systematically by the U.S. Weather Bureau from 1937 onward. By World War II, a network of upper-air sounding stations covered the United States, and the data was the foundation for synoptic meteorology education. Penn State granted its first meteorology Bachelor's degrees in 1942 using weather chart analysis, radiosonde data, and surface observation records as the primary instructional materials. The professorships created in WWII-era training programs were built around teaching students to read these charts and soundings -- a skill that transformed atmospheric science from a vocation for weather hobbyists into a university discipline.

• Mainframe numerical weather prediction1950–1975

  Mainframe numerical weather prediction (IBM 701, ENIAC first forecast 1950)

  On April 5, 1950, Jule Charney and colleagues at the Institute for Advanced Study in Princeton ran the first successful numerical weather prediction on the ENIAC computer -- a round-the-clock forecast computed in a day of machine time. This moment marks the origin of numerical weather prediction (NWP) as a computational discipline, and it transformed the theoretical content of atmospheric science education. By the late 1950s, university professors of meteorology were teaching mathematical models of atmospheric dynamics rather than (or alongside) empirical chart analysis. The IBM 701 became available to universities through the late 1950s, and facilities like NCAR (National Center for Atmospheric Research, founded 1960) gave faculty access to machines capable of research-grade forecasting. Earth science faculty gained access to similar large-scale computing for seismic modeling and oceanographic circulation modeling through the 1960s.

  Effect on the work

  The arrival of computing fundamentally changed the hiring profile for atmospheric and earth science faculty: a PhD in mathematical modeling became increasingly valued alongside traditional observational expertise, and departments began hiring dual-track faculties of field scientists and modelers.

• Weather satellites1957–1980

  Weather satellites and GOES system (Sputnik era; GOES-1 launched 1975)

  The Sputnik launch on October 4, 1957, shocked the US science establishment and directly precipitated the National Defense Education Act of 1958, which expanded NSF funding from $34 million (1957) to $134 million (1959). Earth and atmospheric science departments benefited substantially: new professorships were created, facilities were built, and the satellite observation era began. The TIROS-1 weather satellite launched in April 1960 returned the first photographs of cloud patterns from orbit. GOES-1 (Geostationary Operational Environmental Satellite) launched in 1975 and began providing continuous, real-time satellite imagery of North America -- data that became core curriculum in meteorology and earth science courses. Faculty now had genuinely global observational data for teaching, a transformation as significant for atmospheric science as the telescope had been for astronomy three centuries earlier.

  Effect on the work

  Satellite data fundamentally expanded the scope and complexity of earth-science teaching: ocean surface temperature, land cover change, volcanic plumes, and storm systems became directly observable phenomena in real time rather than reconstructed from sparse surface observations. This expanded the research scope of faculty and drove hiring across all four subdisciplines in the 1960s and 1970s.

• Workstations1975–2000

  Workstations, GIS, and personal computing (NCAR Graphics, ArcInfo, seismic analysis software)

  The arrival of Unix workstations (Sun Microsystems, early 1980s) and the Macintosh (1984) put scientific computing directly on faculty desktops. NCAR developed a suite of scientific graphics and analysis software through the 1970s and 1980s that became standard in atmospheric science courses. ArcInfo (1982) launched the era of desktop GIS, transforming the way geology and earth science courses taught spatial analysis. Seismic analysis software (XSEED, SAC -- Seismic Analysis Code, developed at Lawrence Livermore in the 1980s) gave seismology faculty the ability to process waveform data on desktop machines rather than through centralized computing facilities. The Doppler weather radar network (NEXRAD, completed 1997) gave atmospheric faculty access to high-resolution precipitation and wind data for the first time. By the mid-1990s, internet access to real-time data (model output, satellite imagery, seismic streams) became available through UNIDATA (an Ucar program established 1983), restructuring how faculty designed observational exercises.

• Hubble Space Telescope1990–2015

  Hubble Space Telescope and large-aperture ground observatories (Keck 1990-1996, SDSS 2000)

  The Hubble Space Telescope launched April 24, 1990, and after the mirror correction in 1993, became the most productive astronomical research instrument in history. Keck I achieved first light in 1990 and Keck II in 1996 -- together they were the largest telescopes on Earth and accessible to university astronomers through telescope time allocation. The Sloan Digital Sky Survey (SDSS) launched in 2000 and produced the largest systematic redshift survey of galaxies, making a flood of standardized optical data freely available to faculty and students. These tools created a generation of observational astronomy faculty who could assign research projects using real SDSS data to undergraduate students -- a fundamental change in pedagogy, moving from demonstration to genuine research participation. The Hubble archive and SDSS data portals are now standard teaching tools in university astronomy courses worldwide.

• Cloud geoscience platforms2003–2020

  Cloud geoscience platforms (Google Earth Engine 2010, Copernicus Open Access 2014)

  Google Earth Engine, launched in 2010, placed petabytes of satellite imagery and ML classifiers in a cloud environment freely accessible to university researchers and students in more than a hundred countries. The Copernicus programme (European Space Agency and EU) from 2014 onward made continuous high-resolution land, ocean, and atmosphere data openly available, fundamentally lowering the cost of incorporating real satellite data into earth science courses. Planet Labs (founded 2010) began providing daily global satellite imagery with AI-assisted change detection layers, used in marine and earth science research courses. PhaseNet (2018), a deep-learning seismic phase picker developed at Stanford and SCEC, compressed weeks of manual seismograph review to hours and became freely available for teaching datasets. These tools restructured the faculty role from gatekeeper of scarce data to guide of abundant, AI-processed data flows.

• AI weather2023–present

  AI weather and earth-system foundation models (GraphCast, Microsoft Aurora, FourCastNet, Pangu-Weather)

  In December 2023, Google DeepMind published GraphCast in Science -- a graph neural network weather forecast model that outperforms ECMWF IFS (the previous global NWP benchmark) on 90% of 1,380 test variables across lead times. NVIDIA's FourCastNet generates a global 0.25-degree forecast in seconds. Huawei's Pangu-Weather, published in Nature (2023), matches ECMWF IFS at up to lead times, 10,000 times faster. Microsoft Research's Aurora (May 2024, arXiv), a 1.3-billion-parameter atmospheric foundation model, outperforms GraphCast and Pangu-Weather on multiple benchmarks and can be fine-tuned for weather, air quality, tropical cyclone intensity, solar wind, and ocean waves. For atmospheric, earth, marine, and space science faculty, these tools have arrived simultaneously as research accelerators (accessible without supercomputer access), as course content (teaching students to evaluate when and why AI forecast models fail requires the exact physical intuition the course exists to develop), and as a new professional competency to impart. The Federal Reserve (February 2026) places Life, Physical, and Social Science Teachers in the above-average exposure band for postsecondary AI impact, consistent with the observation that routine lecture preparation, question-bank generation, and structured grading have become substantially compressible -- while field supervision, AI-model critique, and graduate research mentorship remain irreducibly human.

  Effect on the work

  AI earth-science tools are augmenting rather than displacing faculty, but the reorientation is significant: the defining faculty advantage is now the ability to explain why an AI weather model is physically wrong, which requires the same geophysical intuition these courses have always aimed to develop -- but now with AI model critique as the explicit pedagogical objective.