DSC 106 · UC San Diego

Beyond the State Average

What state averages hide, satellite imagery reveals. A county-by-county look at the climate stress shaping Iowa's corn, Kansas's wheat, and Texas's cotton.


Watch the 2-minute demo
scroll to begin

The Setup

Three states feed a nation.

The backbone of a country is its agriculture. The United States, in particular, has some states responsible for a large part of it. AND farmers, and traders all depend on accurate climate data to predict crop turnout and fair market prices. Iowa, Kansas, and Texas grow the corn in your tortilla chips, the wheat in your bread, and the cotton in your t-shirt. Together, they form the engine that fuels American agriculture. However, each state's agriculture is largely affected by the climate in which they grow.

The Conflict

This climate information is extremely important. The turnout of produce affects the price at which it can be sold, and traders in the industry have to constantly keep track of the satellite NDVI to keep up. However, having the same average across the state, is problematic because the entire state does not have the exact same climate, and by grouping the entire region as if it did, leads to an inaccurate estimate of what the produce should cost.

But state averages tell a comforting fiction.

Encompassing the entire state with one number is not accurate to the real climate of a state. Western Kansas is a desert, while Eastern Kansas is humid. Despite this, most climate dashboards will represent this with a single number that ends up not being indicative of the real story.

Inside the average

The big number beside each state is what is typically reported by most dashboards. The colored texture across the state is the actual climate across 458 counties. Hover over any county to read the counties' means and gap from the states' mean.

The Resolution · Our flagship tool

A time machine for America's farms.

THEREFORE, we should built this interactive dashboard to reveal what state averages miss. By combining satellite measurements of vegetation, temperature, and precipitation, the visualization exposes how climate stress unfolds differently across individual counties in Iowa, Kansas, and Texas. Explore seven key points in the agricultural calendar or take control yourself, and hover over any county to see its unique 12-month story emerge.

Scene 1 of 7 · January

The world is asleep.

In winter, the land is dormant across all three states. NDVI hovers near zero everywhere.

January drag, or press play

Iowa

Corn

Kansas

Wheat

Texas

Cotton
Hover any county to reveal its full-year story →

The Lifecycle

Every crop draws its own ring.

Read the year clockwise from January. The further the line from center, the greener the land. Iowa explodes upward as corn matures in July, Kansas climbs steadily and stays high through summer, while Texas peaks early in May and then dips as the climate gets warmer.

Iowa · Corn Kansas · Wheat Texas · Cotton Dashed rings: 0.3 Emergence 0.5 Mid 0.7 Peak Greenness

The Relationships

Heat and rain do not love crops equally.

Each faint background small dot is one county in one month. Each numbered bubbles from 1 to 12 is one state in one month. The line with arrow connecting points is an indicator of time sequence from January to December. By observing what happens when we isolate a single crop, we begin to see three very different stories of the relationship between climate and vegetation.

Filtered by
Plotted against

LST Day Temperature vs Vegetation

Land Surface Temp — Day (°C) on x-axis · NDVI on y-axis

Texas is consistently the hottest every month, but Iowa's more moderate daytime soil temperatures push NDVI the highest. Pure heat doesn't decide greenness. it’s the timing of warmth for emergence and the avoidance of heat stress during peak and harvest.

A Hidden Variable

Crops feel the night too.

Plants respire after dark. The wider the day and night gap in land surface temperature, the more energy crops burn just to stay alive intead of building grain or fiber. Below, each panel splits day-time LST from night-time LST for every month of 2023. (Remember the difference of NDVI trend of cool night state Iowa and warm night state Texas from above scatterplot)

So what did we learn?

Three findings, one warning.

01

State averages can lie by 0.6 NDVI.

In April 2023, eastern Texas counties hit a near-rainforest 0.79 while western Texas counties never broke 0.15. Reporting a single state value masks a 0.64 gap inside one state. This difference is larger than any difference between Iowa, Kansas, and Texas as wholes.

02

Temperature alone does not explain greenness.

Texas is consistently hotter than Iowa, but Iowa's NDVI peaks higher. Crop choice, the timing of rain, and even the night-time chill matter just as much as raw daytime heat. Around 30 °C result in thriving corn in Iowa, who gets plenty of water and less bounfitful cotton across West Texas, which suffers from less water from June to August.

03

The night tells the other half.

Texas night-time LST stays above 20 °C from June through September, denying cotton the cool hours plants use to recover. Resulting in lower NDVI. Iowa's nights, by contrast, average 18 °C even in July which gives the corn peak which can readily be seen.

The warning

Time to retire the state average.

The 0.64 NDVI gap we measured inside the state of Texas in April 2023 is not an outlier, but is instead what every farm state looks like once you stop averaging. As climate change widens the precipitation divide inside each state, that gap will only deepen. State-level reporting was an approximation born of measurement limits decades ago. Those limits no longer exist.

What we built · What was hard

What we have done

  • Pulled MODIS NDVI (MOD13A2), LST day & night (MOD11A2), and CHIRPS precipitation for every county in Iowa, Kansas, and Texas for the full 2023 calendar year, which were then all aggregated to monthly means inside Google Earth Engine.
  • Built an eight-section scrollytelling page in D3.js, anchored by six interactive visualisations: a US intro map, a state-vs-county reveal, a 458-county time machine with sparkline hover, a radial crop calendar, paired scatter plots with crop-stage annotations, and a day-vs-night LST chart.
  • Added a site-wide °C ↔ °F toggle so American and international readers can each read the data in their native unit. This updates every choropleth tick, axis label, tooltip, and inline number on the page re-renders in place, without a page reload.
  • Authored an And-But-Therefore narrative connecting satellite observation to crop-specific lifecycle knowledge.

What was the most challenging

  • Designing interactions that reveal patterns a static plot cannot (especially the per-county sparkline-on-hover), which lets readers compare one county's full-year trajectory against the regional choropleth without leaving the page.
  • Keeping three side-by-side maps (Iowa, Kansas, Texas) visually balanced when Texas is nearly 5× larger than Iowa. This was done by having each map is independently fit to its own bounds while sharing one global color scale.
  • Implementing the °C/°F toggle correctly. Absolute temperatures need the +32 offset (0 °C = 32 °F) but temperature differences only scale (a 21 °C gap is 38 °F, not 70 °F). The two cases live in separate API functions so a tooltip's Day, Night, and Gap rows convert them independently.
  • Writing prose that turns satellite numbers into a story that someone without prior exposure to the data can understand.