When More Records Do Not Mean More Disease

The bias

Raw occurrence counts mix two signals: where a pathogen has been observed and where institutions publish the most biodiversity records. Nearly half of the public Xylella records came from one country, but that did not mean disease pressure was highest there.

The correction

For each country, divide its Xylella records by its total GBIF publishing volume:

$$\text{effort-corrected rate} = \frac{\textit{Xylella}\ \text{records}}{\text{all GBIF records}} \times 10^6$$

This produces a comparable rate per million records instead of a leaderboard dominated by database size.

Surveillance portfolio

Reporting effort reshapes the country ranking

The United States leads raw records but falls to seventh after normalizing for total GBIF publishing volume.

01Iran

7.7

02Italy

7.5

03Brazil

2.2

04Argentina

1.0

05China

0.7

06United Kingdom

0.5

07United States

0.4

08Spain

0.3

09France

0.3

10Taiwan

0.2

Normalized index · Xylella records per one million records published by that country

Evidence detail

Country	Normalized index	Xylella records
Iran	7.7	10
Italy	7.5	104
Brazil	2.2	96
Argentina	1.0	19
China	0.7	10
United Kingdom	0.5	122
United States	0.4	543
Spain	0.3	31
France	0.3	63
Taiwan	0.2	5

The validation step

A normalized index can still miss an outbreak, so it needs an external check. I compared the first public record in each European country with documented plant-health reports.

Data-readiness check

Public data does not provide a consistent outbreak alert

Italy appears on time, Spain within one year, Germany a decade late, and Portugal is missing entirely.

Italy

On time

Spain

+1 year

France

3 years earlier

Germany

+10 years

Portugal

×

Missing

Confirmed outbreakFirst public record2012 → 2026

Evidence detail

Country	Confirmed	First public record	Timing
Italy	2013	2013	On time
Spain	2016	2017	+1 year
France	2015	2012	3 years earlier
Germany	2016	2026	+10 years
Portugal	2019	No record	Missing

Italy appeared in the same year. Portugal's 2019 outbreak never appeared in the occurrence data. That absence is the limit of the method: correction makes comparisons fairer, but it cannot turn incomplete surveillance into proof of absence.

Use this when

• groups differ greatly in how much data they contribute;
• the outcome is a count generated by observation effort;
• an external benchmark exists to test whether the corrected signal is credible.

Do not use it to claim

• true disease prevalence;
• that an empty map is disease-free;
• that every source record has equal diagnostic quality.