Spatial Thinking in Meteorology: A Task Analysis of the Forecasting Process

Weather forecasting, involving the visualization of three-dimensional processes from two-dimensional maps, interpretation of computer-generated graphics and hand plotting of isopleths, draws heavily on spatial thinking. In this study, we used spatial thinking theory informed by cognitive science to investigate the task processing of expert and novice meteorologists with a goal to inform operational meteorology and meteorology education. This is the first step in developing instructional strategies that can scaffold the learning of struggling students in meteorology courses as well as understanding the practices of operational meteorologists.

A pilot survey investigating how meteorologists and meteorology students engage in spatial thinking during forecasting was created to inform the research. The survey was administered to fifty participants at the American Meteorological Society's annual meeting in January 2016 followed by online administration through the spring of 2016. Using written explanations and diagrams, we introduced participants to six types of spatial thinking previously identified in the geoscience literature: visual penetrative ability, perspective taking, mental animation, mental rotation, object location memory and disembedding. Participants then interacted with nine products illustrating a weather event from the fall of 2015, including visible and water vapor satellite imagery, radar base reflectivity and velocity products, surface observation analyses, 500 mb geothermal height plots and model forecast four-panel plots. Finally, participants indicated whether they used each of the six types of spatial thinking in interpreting each product. Data analysis suggested that two specific spatial skills, mental animation and disembedding figure highly in the forecasting process.

Using cognitive science theory as a basis for further investigation, we characterized mental animation as a type of intrinsic (within-object) mental transformation that involves the dynamic visualization of how an object (in this case, a mid-latitude cyclone) changes over time. Similarly, we characterized disembedding as an intrinsic, but static spatial skill that requires isolating and attending to one aspect of a complex display (such as a plotted surface data map). We subsequently investigated how these specific spatial skills, together with domain specific knowledge and working memory capacity affect the ability to produce an accurate forecast. Expert and novice meteorologists (as measured by a domain experience questionnaire) completed psychometric tests of their spatial thinking ability including the revised Purdue Spatial Visualization Test (Visualization of Rotations) and the Educational Testing Services’ Hidden Figures Test. Because previous studies suggest that weather forecasting depends largely on strategies designed to reduce working memory load, we also used matrix and arrow span tasks as measures of participants’ working memory capacities. Finally, the expert and novice meteorologists completed a meteorology concept inventory developed and tested by collaborators at the United States Air Force Academy during the fall semesters of 2013 and 2014.

To evaluate participants’ ability to produce an accurate forecast, we provided a surface chart and three upper air charts depicting the same mid-latitude cyclone. Participants were asked to mark the low-pressure center, annotate troughs and ridges, draw in warm and cold fronts, shade areas of cold and warm air advection, positive and negative vorticity, divergence and convergence and predict the location of lowest pressure twelve hours in the future. The resulting participant products were scanned, digitized and compared to a key for accuracy using a geospatial processing program (ArcGIS) that determined the percent of pixels in a participant’s map identified as correct. A regression analysis of the collected data investigated the effect of the predictor variables on the outcome task. A think-aloud protocol with selected participants provided a qualitative look at processes such as task decomposition, rule-based reasoning and the formation of mental models in an attempt to understand how individuals process this complex data and describe outcomes of particular meteorological scenarios. With our preliminary results we aim to inform operational meteorology and meteorology education from a cognitive science perspective. The results point to a need to collaborate with the atmospheric science community broadly, such that multiple educational pipelines are affected including university meteorology courses for majors and non-majors, military weather forecaster preparation and professional training for operational meteorologists, thus improving student learning and the continued development of the current and future workforce.