"The use of automated data collection, mining and analysis for future farm management"

Transcription

1 "The use of automated data collection, mining and analysis for future farm management" G. Katz, A. Arazi * S.A.E. Afiimilk, Kibbutz Afikim, Israel

2 Outline Introduction Automated Data collection Data mining and analysis Summary 2

3 Data Collection, mining and analysis Data Information Knowledge Intelligence Optimization Predictive Modeling What is the best that could happen? Raw Data Standard Processed Reports Data Descriptive Modeling Analytical on-line Reports What happened? Why did it happen? What is going to happen? 3

4 Real Time acquisition of milk components, yield and conductivity Milk meter Milk analyzer * Free flow * Non-interfering measurement * Continuous real time acquisition of milk components * Data is acquired automatically for the individual cow during its milking 4

5 Behavior Sensor (pedometer + ) How do we know how an animal feels? Assuming an animal manifests its feeling by its behavior there are numerous aspects of behavior Activity Moving steps Lying Lying time Lying bouts Calmness Restlessness (option) Pedometer Plus measures: Activity Steps Rest time Minutes Rest bout - # Antenna visits # - Calmness/Restlessness 5

6 Scheme for a Data Recording and Management System Sensors Row Data Milk quantum and flow Milk conductivity Milk components Database & Management Decisions & Reports Cows production Health Milking efficiency Estrus detection Devices and experts Milking parlor Cow activity Cow behavior Cow weight Individual feeding Milk separation Energy balance Calving detection Cows welfare Dairy experts 6

7 Data mining and analysis Major Goal: Maximize Milk yield and Quality and minimize the Cost of production Primary Factors production Feed ration Health and welfare Reproduction

8 Milk Payment ECM Milk production, %Fat, %Protein, SCC ECM (NIS) Schcolnik et. al, 2007

9 Fat (%) Nutritional Information - Control Nutritional Status Feed ration Consumed ration On-Line milk component data rapid detection of nutritional changes Nutritional problems > fat depression, Fat/Protein fluctuations -> energy efficiency Or moldy feed Schcolnik et. al, 2007 Protein (%)

10 Nutritional Information - Individual Feeding igh importance Management where supplement of additional concentrate feeding is needed (pasture, robotic milking, fresh cows) RC 2001 formula: Daily data required for calculation: Fat Protein Milk yield Body weight Week of lactation MI (kg/day) =(0.372*FCM *BW)*(1-e *(wol+3.67) )

11 Individual Feeding - Example Cow 2823 Allocation concentrate feeding according to 4% Fat E. Maltz, personal communication Over estimation of 60 kg concentrates by periodic test most likely leading to further decline in fat

12 Mastitis Control Mastitis Case Report Commercial Kibbutz Farm Correlation between udder health and milk components Lactose and conductivity two of the most promising parameters (Pyorala, 2003) Lactose SCC Yield Conductivity

13 Mastitis Control Mastitis Case Report Commercial Kibbutz Farm Lactose SCC Yield Conductivity

14 Cow Health Monitoring Early Mastitis Detection Behavior changes significantly during one day prior to milk drop and Daily milk yield (kg) Milk conductivity sharp increase in milk Days Daily milk yield Milk conductivity conductivity (mastitis diagnosed in day 7) No. of steps Lying time (min) days Number of steps Lying time (min) 14

15 Clinical Mastitis Event Yield Lying time conductivity SCC

16 Abdominal Pain Lying bouts Lying time Ѵ

17 Eye Clinical Scratching Injury Mastitis Event Lying time Ѵ Ѵ activity

18 Cow Health Monitoring Diagnosis of Metabolic Diseases Correlation between metabolic diseases and milk components Ketosis (NEB) Fat/Protein Ratio (FPR)> (Heuer et. al., 1999) SARA FPR< 1.0 or more then 10% with fat<2.5% (Tomaszewski and Cannon, 1993; Nordlund et. al., 2004)

19 Diagnosis of Ketosis FPR FPR BHBA>1.4 (31.3%*) Sensitivity (%) AfiLab (Laboratory) Specificity (%) AfiLab (Laboratory) > (90.3) 56.1 (37.4) > (45.2) 82.7 (75.5) > (25.8) 92.4 (92.8) > (6.5) 98.3 (97.8) * % of cases with BHBA above threshold Schcolnik et. al., 2007

20 Diagnosis of Ketosis Multifactorial Approach FPR cut off FPR + SHI 1 filters + BHI 2 filters Sensitivity (%) Specificity (%) SHI Sound Health Indicators 2 BHI Bad Health Indicators Livshin et. al., 2008

21 Ketosis Case Report Ketosis events

22 Suspected SARA by Group

23 Fat concentration fluctuations between milking sessions mean sd = 0.57 fat%, mean peak-to-peak = 2.16 fat % Comparing lab results to on line Analyzer for a typical Holstein cow Max Peak to peak fluctuation for all cows during experiment session The range of our ensemble for a given 10 days period 30 consecutive milking sessions in 10 consecutive days was sampled in the lab and by the analyzer (A.R.O farm, n=88 Holstein cows).

24 Feeding Management %Fat by Milking Difference of Fat between Milking (Group level) % Fat % Fat <2.5% Ratio Fat:Pro t Ratio

25 Feeding Management %Fat by Milking FAT% Fat/Prot Morning noon evening

26 Heat Detection Cows in heat are usually restless (walk more and lie down less) Restlessness Heat detection through Milk yield changes in lying behavior Fat Lying time Activity

27 Normal Heat Behavior

28 A c t i v i t y Silence Heat

29 Detecting Calving Time Helpful tool for daily routine plan Attend expected difficult calving Cows behavior changes prior to calving

30 Detecting Calving Time Activity/Lying ratio in the last 7 days prior to calving * Ѵ * P< 0.01 day -1 vs. day -2 Maltz and Antler, 2008 A.R.O

31 Applied Research Development and Cooperation

32 S.A.E. Afikim Applied Research Cow health monitoring Afikim farm Heat detection free stalls barns Cooperation with Volcani Center (Israel) Meimad farm Welfare and Comfort group level Deviation parameters for Pasture management South Africa Pasture quality and availability information

33 S.A.E. Afikim Applied Research Team Improving calving time detection Cooperation with Volcani Center (Israel) Heat detection in and management Commercial farms - Germany and Poland Improve cow health and fertility monitoring specification and timing Integrated data Self feeding allocation Afikim farm Crazy Elephants?!!! Jerusalem Zoo

34 Academic Cooperation uelph University (Canada) arly diagnosis of lameness cows arly detection of cows suspected for postpartum diseases - cooperation with Volcani Center

35 Academic Cooperation Oregon State University (U.S.A.) Welfare and comfort individual cow level Detect health problems and abortions I.R.T.A. (Spain) Effects of group changing behavior and production

36 Academic Cooperation Volcani Center (Israel) Effect of group changing Behavior, fertility and production Virginia tec(usa) Genetic evaluation, Udder health University of Florida(USA) Economic decision making in farm management

37 Thank you for your attention