Kamphuis & huijps, ECPLF, milan, 2015 ss

Post on 21-Feb-2017

164 views 0 download

Transcript of Kamphuis & huijps, ECPLF, milan, 2015 ss

Using successful inseminations to evaluate

sensitivity of automated heat detection systems

Claudia Kamphuis, WURKirsten Huijps, CRV

Challenges of detection-Time consuming-Error prone-Increased herd sizes

The importance of heat detection

New Zealand survey 500 farmers25% wants it7% has it70% listed it in top 3 of technologies that gained benefit for farm(Edwards et al., 2014, APS)

Survey 212 farmers stating to have sensor technology41% of AMS farmers has it70% of CMS farmers has it(Steeneveld and Hogeveen, 2015, JDS)

Survey 109 farmers globally41% has itRated as useful to very useful(Borchers and Bewley, 2015, JDS)

35% of US respondents(Bewley, 2014, EAAP/EU-PLF conference)

Adoption rates

What are the success factors?

Associated with clear management actions

What are the success factors?

Associated with clear management actions

Investment is economically beneficial (Rutten et al., 2014)

for Dutch circumstances

What are the success factors?

Associated with clear management actions

Investment is economically beneficial (Rutten et al., 2014)

for Dutch circumstances

Sensitivity 80% with specificity 95% (Rutten et al., 2013, JDS)

performance is OK

But a specificity of 95%....

Still requires visual confirmation

701-05-2023

But a specificity of 95%....

Still requires visual confirmation

100 cow herd:~1 cow in heat / day99 cows not in heat / day5 falsely alerted cows / day

801-05-2023

Alerts around P4 determined heat events(Kamphuis et al., 2014, EAAP/EU-PLF, Copenhagen, Denmark)

84% of all alerts generated +/- 3 days around P4 heatOnly 30 P4 heats included

Use successful inseminations

Avoid high costs and intensiveness of P4 method

Relatively easy way to increase numbers

Additional: study differences in performance between parities and lactation stages

Data collection: January-July 2014

Insemination data from CRVHeat alert data from systems’ software

Farm A Farm B

Herd size 450 250

Milking system Conventional AMS

Milk production (kg/cow/year)

9,500 9,800

Heat detection system A and B B and C

Defining successful inseminations

For cows with just 1 insemination

31 July1 January

Insemination>56 d

Successful GS+

Defining successful inseminations

For cows with >1 insemination

Insemination

>56 d

Successful GS+

Insemination

≤56 d

Unsuccessful

31 July1 January

Insemination

≤ 56 d

Unsuccessful

Insemination

>56 d

SuccessfulGS+

Insemination

≤56 d

Unsuccessful

1 January 31 July

Matching inseminations with alerts

Insemination data at daily level

Heat alerts per h or per 2h time blocks‘summarized at daily level’≥1 alert per day is an alerted day

Data collection: January-July 2014

Farm A Farm B

Herd size 450 250

Milking system Conventional AMS

Milk production (kg/cow/year)

9,500 9,800

Heat detection system A and B B and C

Cows with ≥1 successful inseminations

145 119

GS positive inseminations

153 129

Heat alerts A: 352B: 532 B: 117

C: 887

Matching inseminations with alerts

Sensitivity using a 1 day time-window

GS+

day

Matching inseminations with alerts

Sensitivity using a 2 day time-window

GS+

day

Detection performance differed between heat detection system

Sensitivity (%)A: 31B: 58C: 92

Detection performance differed between heat detection system

Sensitivity (%)A: 35B: 87C: 94

Detection performance did not differ between parity / lactation stage

Sensitivity 2 day time- window

System A System B System C

Parity

1 23 85 90

2 48 90 97

≥3 36 90 94

Lactation stage

≤56 days 44 100 91

>56 days 34 87 94

Discussion / conclusion

Lower performance of System A• consistent with EAAP/EU-PLF 2014• technical problems in data exchange

Discussion / conclusion

Lower performance of System A

Sensitivity may be overestimatedSummarizing h / 2h alerts to daily alertsexcluding data from cows that were

not detected by a systemnot visually confirmed by farmerinseminated incorrectlynot inseminated at all

Discussion / conclusion

Sensitivity in line with previous research (Rutten et al., JDS, 2013)

around 80-90% using a 2 day time-window

System C also alerts often 1 day after a successful insemination

No apparent differences in parity/lactation stageno apparent need for specific algorithmsfocus on getting more accurate and precise alerts

Ongoing data collection

Confirm or refute current findings statistically

Analyze data using hourly or 2hourly alert basis

Where to from here

Thank you for your attention

www.slideshare.net/claudiakamphuis ckamphuis