1 Arbel et al. (2001) JDS 84: 600 Preliminary conditions changed? - PDF document

Lecture outline Advanced Herd Management KU-Life 23-10-07 1. Presentation of the problem 2. Choice of method (Monte Carlo Sim ulation) 3. Application of the method Economic consequences of postponed first insemination of 4. Results cows in dairy herds -An application of Monte Carlo simulation 5. Evaluation of the applied method Søren Østergaard, Aarhus Universitet, Faculty of Agricultural Sciences, Department of Animal Health, Welfare and Nutrition A A R H U S U N I V E R S I T E T Det Jordbrugsvidenskabelige Fakultet 1. Presentation of the technical problem Prolonged lactation � Example: The application of SimHerd for the calculations of consequences due to Longer calving intervals prolonged lactation Later insemination start - planned! � Educational example: Application of the Monte Carlo simulation - Fewer calves � Changing the first drawn conclusions from the + Fewer diseases - Longer late lactation test data + Fewer days being dry (equals lower yield) Economical loss – Danish kroner per Typical reproduction management empty day � 4,6 - 7,6 Olds et al. (1979) Empty days are expensive � 3,0 - 22 Pedersen (1981) Many exam inations � 21 - 26 Bailie (1982) 0 - 25 Danish kroner per day � 2,1 - 7,1 Dijkhuizen et al. (1985) � 10,2 Reproduction objective Weaver & Goorger (1987) � 0 - 3,9 Schm idt (1989) 90 - 100 empty days in average � 0,3 - 12 Strandberg & Oltenacu (1990) Early insemination start (40 days)! 1

Arbel et al. (2001) – JDS 84: 600 Preliminary conditions changed? 1008 high-yielding cows from 19 herds More persistent lactation curves High yield and low body condition score 1. calf Older at drying off Ins. start 90 & 150 60 & 120 Lower meat/ milk price proportion Empty days 128 & 189 110 & 160 Milk 1-10 months +0,9% +2,0% Higher focus on diseases Milk 1-5 following +4,1% +0,9% months Economic effect +5,2% +3,4% Financial analysis employed by Financial analysis employed by Arbel et al. Arbel et al. � Income � Problem s � Milk (T> C) � In practice: Will the number of culled cows, calves born � Culled cows (Test design!; NS) and heifers in calf remain similar? � Calves born (NS) � Is the higher yield associated with the feed � Expenses consumption? � Feed consumption (Standard numbers for DIM; T> C) � If the insemination start is postponed, how will the � Purchased pregnant heifers (one per culled cow; NS) actual distribution of calving intervals be? � Work expenditure (Fixed number per annual cow; NS) � Will it differ between herds with different reproduction efficiency? � Period � Does the chosen study period reflect the herd result? � 1½ lactation; Same number of days for treatment and control Presentation of the technical … the problems – as to principles problem – summary: � The consequences of the postponed � Cow level vs. herd level insemination start calculated at the herd � The state of a cow depend on: level based on new estimates at the cow � The former state level � Transition probability to the present state (stochastic) � Herd production strategy (between herd variation) � Condition of the rest of the herd (mechanistic system) 2

The history of SimHerds 2. Justification of method � Monte Carlo simulation is chosen because: � Developped by Jan Tind Sørensen (DJF) in 1989-91 � A new recommendation based on experimental test � Scientific tool results not including essential correlations at herd level � Advisory tool interplaying with HerdView in 1991-92 � Farm studies demand many farms performing the test � SimHerd II including a disease module in 1998 design over a longer period (very expensive) � SimHerd III including an even more detailed disease � SimHerd is chosen because it already exists module in 2003 � ’Possible’ alternatives � SimHerd IV including more details on mastitis � Partial budgeting (Stochastic) � Dynamic programming � Different special editions regularly updated Applications of the SimHerd, examples: How does the model function? � Influence of the sickness level (1988) � Mortality and disease occurrences (1991) � SimHerd I + strategies of feeding, reproduction and renewal (1992) � Simulating the technical and econom ical � Length of empty period (1993) � Strategies of renewal and reproduction (1995) consequences of the production strategies in a � BVD (1995) dairy cattle herd � Application of the TMR-1 (1996) � Grazing intensity (1997) � Organic vs. conventional production (1998) � Time for the first insemination (1998) � ’technical’ – e.g. m ilk yield and disease incidence � Strategies of combating the staphylococcus mastitis (1999) � SimHerd II + interplaying with the feed, health and production � ’econom ical’ – e.g. DB per annual yield of a cow (2000) � Application of a pen bull (2002) � ’production strategy’ – e.g. increased heat � AMS and renewal (2002) detection � Prolonged lactation (2003) � SimHerd III + Strategies of controlling milk fever (2004) � Strategies of controlling the PTB (2004) � Information value of progesterone measurements (2004) � Early treatment of mastitis and ketose (2004) � Economical values of breeding attributes (2004) Input og output The design of SimHerd Input: � State variables � Dynam ic � List of all the cows and heifers in the herd with � Simulation of the development in herds over time parameters for their current state at simulation start � Time step of one week � Parameter values for the model � Mechanical � Biological parameters � Simulation of the yearly results of the herd via ’parallel’ � Parameters describing the production system (e.g. barn capacity simulation of the single animals � Parameters describing the production strategies (e.g. � Stochastic feedstuff plan, reproduction and culling strategy) � Variation between the cows’ production capacity and the Output: events on cow level simulated with random numbers and � Technical annual results relevant probability distribution � Typically 10 years from 10 to 1000 repetitions 3

Why is SimHerd useful The simulation steps – generally speaking? � A cow in each state � The choice of production strategies is of great � Lactation stage econom ical im portance for the farmer � Reproduction state (Heat, gestation, abortion, calving) � Disease, death, voluntary and involuntary culling � Calculation of the consequences of the production � Feed intake strategies is complicated if done with exactitude � Milk yield and growth � Interaction between the employed strategies � A heifer in each stage � Age, reproduction, culling, feeding Effect on cow level × number of cows � At herd level in each state ≠ � Replacement (max. number of cows, cows on culling list, heifer in calf, strategy of buying and selling heifers) Effect on herd level � At the end of a sim ulation year, the results are saved on a file … continued 3. Describing the procedure � ”All things being equal” obtained via the model � In practice the production strategies will alter within the given test period I. Form ulation of strategies/ scenarios II. Parameter estimation � Sim Herd can serve as a educational tool • If necessary based on published test results demonstrating correlations at herd level III. Programm ing (if required) IV. Validation � SimHerd can point out the missing knowledge and • Internal (testing if the programme is doing as it is calculate the consequences supposed to) • External (Face validation; sensitivity analysis) V. Sim ulation of experiment VI. Analysis and interpretation of results Optimise the complexity Pressure for an increased degree of detail � We can predict scenarios with a combination of SimHerd and external � ”But there is a correlation” � calculations ”We should include it for safety reason” BUT � Additional state variables are often very � Increased complexity makes the validation more expensive difficult! � ’face validation’ is typically the only possibility as an � SimHerd still being used might indicate a external validation mothod respect for complexity � Influence on the present form of SimHerd � Validation is turning problematic 4

Scenarios with prolonged lactation 4. Results - Simulated using SimHerd 1 2 3 Short Long, 1st Long, all Ins. start, days 1st calf 70 140 140 Older 35 35 105 Scenarios with prolonged lactation Two herds 1st calf Older Reproduction management Milk 1-10 months +1,0 % +1,8 % Average Good Milk 1-5 months +0,9 % ! +0,7 % following Insemination chance % 45 60 Conception chance % 50 50 Ins. period – high yielding 189 175 Ins. period – low yielding 147 133 Unchanged risk of diseases, reproduction, feeding etc. Fewer culled cows A typical Danish herd � 120 cows 44 � Own breed � Yielding level approx. 8500 42 � Feeding includes 3 total m ixed 40 Repro Udskift. % rations 38 Middel � Dry period of 7 weeks God 36 Min. yield of 12 kg � 18 - 20 % involuntary culling 34 � Calf m ortality from 8% to 12% 32 � Typical disease occurrences 30 Tidlig alle Udskudt 1. Udskudt alle Ins. start 5