Four ways to render one field

The same yield data classified four ways, where equal-interval collapses almost the entire field into a single class

Academic coursework  /  January 2022

Discipline
Cartographic judgement
Tools
ArcGIS
Methods
equal interval classification, quantile classification, natural breaks classification, standard deviation classification
Data
Manitoba Zero Tillage Research Association; Assiniboine Community College

Notes on method

This is the plate I would show first, because it is the only one that argues about method rather than reporting a result.

Every panel holds the same points, the same attribute, and the same colour ramp. Only the class breaks move. Equal interval divides the range into five equal slices, and because a handful of points reach 306 bu/ac while most of the field sits below 60, the first slice swallows almost the entire field. The map is not wrong. It is correctly rendering a scheme that the distribution defeats.

Quantile forces equal counts into each class, which maximises visual contrast and produces a field that looks violently variable. Natural breaks finds the gaps the data actually has. Standard deviation reframes the question entirely: it stops asking what the yield is and starts asking how far it sits from the mean, which turns out to be not very far across most of the field.

A reader handed any one of these alone would come away with a different impression of the same harvest. That is the argument. The classification is not a display setting applied after the analysis; it is part of the claim being made.

Four panels of the same field, each classifying dry yield differently: equal interval renders almost the whole field dark green in the lowest class, while quantile fills it with reds and greens, natural breaks shows banded structure, and standard deviation shows most of the field within half a deviation of the mean.
One field, one attribute, four classification schemes. Equal interval (top left) puts nearly every point in the lowest class because the range runs to 306 bu/ac while most values sit under 60.Base data: MZTRA and ACC (2006). Scale 1:5,000Open larger (2432px)