Lecture – Physiology basics → what data must answer
ANSCI 4040 – Dairy Data Analysis & Analytics
Ass. Prof. Dr. Miel Hostens
Learning goals
Refresh core reproductive physiology (enough to interpret data)
Translate biology into questions your database must answer
Identify common failure modes (estrus, ovulation, luteolysis, anovulation)
Connect management levers to measurable outcomes
This deck is intentionally “physiology-light but analytics-heavy”. We focus on the minimum biology needed to interpret KPIs correctly.
Why repro is a data problem
Reproductive efficiency drives:
Milk production dynamics (lactation length, calving pattern)
Replacement rate, culling risk, heifer inventory, genetic progress
Labor and treatment costs
The industry’s challenge: fertility declined then rebounded—requiring integration of physiology, management, nutrition, genetics, economics, and records.
Conceptual model: “open → pregnant”
How fast the program turns eligible non-pregnant cows into pregnant cows.
Key workflow:
Conceptual model: “open → pregnant”
Minimal physiology you must know
Estrous cycle in cattle:
Minimal physiology you must know
Six key hormones and what they mean for data interpretation:
Progesterone (P4) : from CL; maintains pregnancy; blocks estrus / GnRH surge
Estradiol (E2) : from follicle; standing heat; triggers GnRH surge
GnRH : hypothalamus; triggers LH surge
Minimal physiology you must know
LH : surge causes ovulation; pulses drive dominant follicle growth
FSH : initiates follicular wave
PGF2α : uterus; regresses CL (luteolysis)
What high milk production changes (and why metrics mislead)
High milk production is associated with:
Larger ovulatory follicles but lower circulating E2
Larger luteal tissue volume but lower circulating P4
Shorter estrus duration at higher production
Higher risk of double ovulation (twinning risk)
Analytics consequence: “heat detection looks worse” partly because estrus expression is shorter , not only because labor/observation is worse.
Estrus: behavior vs biology
Standing heat is driven by E2 and ends when E2 drops below threshold (often before ovulation).
AI timing should relate to onset of estrus (biologically tied to GnRH/LH surge), not just “was she seen?”
Therefore: “no observed heat” ≠ “no ovulation”
Anovulation: big hidden driver
Anovulation commonly defined as failure to ovulate around ~60–80 DIM
Multi-study summaries show ~20–25% anovular cows around 60–70 DIM (large between-farm variation).
Two phenotypes (useful for interpreting progesterone/ultrasound data):
“Small-follicle” anovular (insufficient LH pulses → follicles don’t reach ovulatory size)
“Large-follicle/cystic” anovular (high E2 fails to induce GnRH/LH surge) 1
Management lever: timed AI programs
Why TAI exists:
Detection-of-estrus-only programs often delay first service and reduce service rate.
Ovsynch (GnRH → PGF2α → GnRH → TAI) coordinates ovulation and CL regression, enabling timed insemination without estrus detection.
Presynchronization (PGF2α and/or GnRH combinations) aims to start protocols at an optimal cycle stage and address anovulation.
“What questions must the data answer?”
For any herd (and each pen/group):
Eligibility / exposure
Who is eligible to be bred this week (DIM, VWP, health holds)?
Submission
Who was submitted to insemination (observed heat or protocol)?
Service quality
Was the insemination recorded correctly (date/time, technician, semen, bull)?
“What questions must the data answer?”
Fertility outcome
Who conceived (preg check), who didn’t, who lost?
Timeliness
How long did it take from calving → first service → conception?
Bottleneck diagnosis
Is low pregnancy rate due to service rate or P/AI ?
Take-home summary
Physiology gives you mechanisms ; metrics give you signals
High production alters hormone dynamics → affects estrus expression and twinning risk
Anovulation is common around breeding start and requires protocol-aware interpretation
Your monitoring framework must separate:
Exposure (service/submission) from success (conception/pregnancy)