Changing methodologies in solar physics

TL;DR

This paper analyzes the evolution of research methods in solar physics over eight decades, examining how technological and environmental changes have influenced scientific training and progress.

Contribution

It provides a historical analysis of methodological changes in solar physics research using PhD theses, highlighting trends and the impact of technological advances.

Findings

Methods have changed little aside from technological advances.

Impactful scientific advances appear to be becoming rarer.

Historical events influence research trends and method adoption.

Abstract

This study attempts to establish a basis for understanding how methods used in research in solar physics have evolved since World War II (WWII). The goal is to begin to explore if and how the changing research environment affects the training of young scientists, and the future of solar physics research at our institutions. A strategy based upon a sample of 650 PhD theses is used to seek possible trends over 8 decades, with the aim of uncovering any correlations between methods used and measures of success. Necessarily subjective, results depend on how methods are defined, and how success is measured. Although a brief justification of the choices made is attempted, trying mainly to avoid pitfalls such as counting citations, it is clear that further assessment is required. The statistical analysis is based upon necessarily subjective categorization and the inference of likelihoods of two different distributions being drawn from the same underlying distribution. The statistics seem to reflect historical events, such as the Kennedy Moonshot program and the associated SKYLAB mission, with changes delayed by a few years. The data suggest that impactful advances are becoming more rare. Yet the methods used have changed little barring those related to obvious technological advances (e.g. the advent of spacecraft, adaptive optics). A follow-up study to explore the 100,000+ publications in solar physics through machine learning seems warranted.