Introduction

A large number of cities in the United States experience variable weather. Weather stations and services attempt to give accurate reporting to the next day’s actual temperature, but in any case, error still exists. In this report I use data concerning regular weather forecasts in 167 cities in the United States and its territories spanning between January 2021 and June 2022 to determine what environmental factors may play a role in weather forecast error.

Results

Of the environmental data included in my data, I interest myself most for how precipitation plays a role in incorrect data forecasting. Sudden storms can cause temperature drops, so it would be logical to assume that some error may be as a result of rain’s cooling effect, which can be complex to measure due to the effect of humidity.1 Plot (a) below is a 2D histogram where instances of of forecast error are plotted on the average annual precipitation of the region:

The groups of cities are not necessarily evenly-spread within this 2D histogram. The amount of rainfall is within itself is skewing the graph towards the left, since there are fewer cities populating the left side of the graph, and the one city with much more rainfall (Juneau, Alaska) does not affect the distribution much on its own. This plot raises its own questions: namely, what do the cities have in common? For this reason I created plot (b), which designates each city by its Köppen Climate Category. As one would expect, arid cities experience far less rain and make up most of the left skew in the plots. It would also seem that cities in each climate category cover a large portion of the y-axis. Based on the fact that cities of continental climate generally have the largest instances of error, with a number of arid and temperate cities less erroneous than the lest erroneus continental city, one might think that continental cities have the least predictable weather. However, these plots are hardly confirmation. This is cause for further investigation, the boxplot below opens room for explanation:

The boxplot above captures the distributions of average forecast error by city, separated by Köppen Climate Category. Note that the scatterplot above counted all instances of error, while this boxplot averages the actual value of the error. It would seem that tropical cities have lower error overall. First of all, only six tropical cities are represented in this data, and second of all, tropical weather has been characterized as “quite predictable”.2 The other side of this is the data concerning temperate cities. There are a large number of outliers in this category, meaning that the weather in this climate type would easily be the most unpredictable. It would seem that arid cities also are difficult to predict, given the wide range of average forecast error, but they all fall within what is a “normal” amount of error (no outliers). This contradicts the natural line of thinking that would arise from the plots above-although there were more instances of error in continental cities, their forecasts are on average incorrect by more similar amounts than in temparate cities. The one outlier continental city is Fairbanks, Alaska. I would be interested in further research regarding this city, as it possesses an average count of error instances, and is in the bottom third for average precipitation. It would be interesting to uncover what it is about this city in particular that makes its weather so unpredictable.

Conclusion

Köppen Climate Categories seem to be a great indicator for variability in the realm of weather prediction. Knowing the climate category of a given city may allow us to be able to make predictions of the reliability of a forecast. The Midwest region, for example, is known for its difficult-to-predict weather, and is itself largely temperate. It would also seem that the amount of rainfall an area receives affects the accuracy of a forecast, though only very slightly.

  1. Liu, Wei, Shaorou Dong, Jing Zheng, Chang Liu, Chunlin Wang, Wei Shangguan, Yajie Zhang, and Yu Zhang. “Quantifying the Rainfall Cooling Effect: The Importance of Relative Humidity in Guangdong, South China”, Journal of Hydrometeorology 23, 6 (2022): 875-889, doi: https://doi.org/10.1175/JHM-D-21-0155.1, 875↩︎

  2. Sobel, A. H. (2012) Tropical Weather. Nature Education Knowledge 3(12):2↩︎