Sushil Paudyal
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Sushil Paudyal has a total of 141 co-authors. Here are the top co-authors by number of shared publications:
- Pablo Pinedo
Department of Animal Sciences, Colorado State University
13 shared publications Rajesh Neupane
Department of Animal Science, Texas A&M University, College Station, USA
7 shared publications- Bhuwan Shrestha
Department of Animal Science, Texas A&M University, College Station, TX 77843
5 shared publications D. Manriquez
Department of Animal Sciences, Colorado State University, 350 W. Pitkin St., Fort Collins, CO 80523, USA
4 shared publications- Juan Manuel Piñeiro
Department of Animal Science, Texas A&M University, College Station, TX 77843
4 shared publications A. Donovan
Department of Large Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville 32611
3 shared publications
Latest Publications
- Open Access
- 1 October 2026
- Published by American Dairy Science Association in Journal of Dairy Science
Strategies for dairy farm business resilience in small and mid-sized US farms
Small and mid-sized dairy farms in the United States face increasing challenges due to economic pressures, environmental variability, and industry consolidation. These farms often struggle with low profitability, high input costs, labor shortages, and volatile milk prices, which have led to a significant decline in their numbers over the past 2 decades. Despite these challenges, some farms have demonstrated resilience by adopting innovative strategies such as automation, organic certification, raw milk marketing, and diversification into beef production. This paper explores the financial and operational difficulties faced by modest-sized dairy farms and evaluates the strategies that have enabled some to remain viable. It also highlights the need for comprehensive decision-support tools that can help farmers assess their financial health and identify suitable management practices. This study uses an explicit, conditions-based decision framework that states, for each strategy, the herd-size, equity, and market-access conditions under which it is the appropriate choice, and consolidates the critical financial thresholds referenced throughout the paper into a single summary. By evaluating financial analysis, this work establishes a framework that empowers farmers to navigate economic volatility, thereby enhancing their operational resilience and long-term sustainability.
- 1 October 2026
- Published by American Dairy Science Association in Journal of Dairy Science
Open AccessStrategies for dairy farm business resilience in small and mid-sized US farms
Small and mid-sized dairy farms in the United States face increasing challenges due to economic pressures, environmental variability, and industry consolidation. These farms often struggle with low profitability, high input costs, labor shortages, and volatile milk prices, which have led to a significant decline in their numbers over the past 2 decades. Despite these challenges, some farms have demonstrated resilience by adopting innovative strategies such as automation, organic certification, raw milk marketing, and diversification into beef production. This paper explores the financial and operational difficulties faced by modest-sized dairy farms and evaluates the strategies that have enabled some to remain viable. It also highlights the need for comprehensive decision-support tools that can help farmers assess their financial health and identify suitable management practices. This study uses an explicit, conditions-based decision framework that states, for each strategy, the herd-size, equity, and market-access conditions under which it is the appropriate choice, and consolidates the critical financial thresholds referenced throughout the paper into a single summary. By evaluating financial analysis, this work establishes a framework that empowers farmers to navigate economic volatility, thereby enhancing their operational resilience and long-term sustainability.
- Open Access
- 12 August 2026
- Published by Elsevier in Smart Agricultural Technology
- Vol. 15
Investigation of poultry feces removal rate and vision-based feces weight estimation on grooved-floor panels
- 12 August 2026
- Published by Elsevier in Smart Agricultural Technology
- Vol. 15
Open AccessInvestigation of poultry feces removal rate and vision-based feces weight estimation on grooved-floor panels
- Open Access
- 1 August 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (8) , 8424-8434
Characterization of failed milking events in a voluntary milking system
Robotic milking systems are increasingly used on dairy farms, offering automation and enhanced efficiency in the milking process. However, failed milking events are a serious concern for overall farm milking efficiency. The objective was to evaluate failed milking events (FME) in a voluntary milking system with free flow cow traffic, to understand impacts and outcomes based on teat positions, lactation stages, parity, and environmental variations. Cow and milking records data from the milking robots were collected from a dairy farm in Texas between February 2020 and June 2024, encompassing a total of 5,053 failed milking events out of 614,681 total milking sesssion. A generalized linear model with a binomial distribution and a logit link function was used to assess the impact of the temperature-humidity index (THI) levels, milk yield, milk speed, cow parity, and cow lactation stage variables on the odds of an FME. Similarly, a generalized linear model with a truncated negative binomial distribution and a log link function was used to evaluate the incidence rate ratio (IRR) of FME across lactation weeks and parity. Overall, teat-connection-related failed milking events were primarily due to connection time (CT; 35.1%) and increased teat-connection attempts (CA; 21.9%) . Dead milk time (DT) and automatic robot stop (AS) contributed 35.4% and 0.05% of total FME, respectively, reflecting the distribution of failed milking reasons . Overall FME rates decreased significantly over lactation weeks (IRR = 0.98, 95% CI: 0.96-0.99) as identified by IRR model. Cows with an impending mastitis event within 7 d had significantly higher odds (OR = 6.70, 95% CI: 4.59-9.75) of FME as compared with healthy cows. Longer intervals between milking significantly increased the odds of FME (OR = 1.79, 95% CI: 1.54-2.08) as compared with regular milking intervals. Milking failure risk declined across lactation in primiparous cows but increased slightly in multiparous cows, demonstrating a significant parity-dependent effect of lactation stage. This information could be used to tailor milking strategies to meet the animals' physiological requirements, and enhance robotic milking efficiency.
- 1 August 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (8) , 8424-8434
Open AccessCharacterization of failed milking events in a voluntary milking system
Robotic milking systems are increasingly used on dairy farms, offering automation and enhanced efficiency in the milking process. However, failed milking events are a serious concern for overall farm milking efficiency. The objective was to evaluate failed milking events (FME) in a voluntary milking system with free flow cow traffic, to understand impacts and outcomes based on teat positions, lactation stages, parity, and environmental variations. Cow and milking records data from the milking robots were collected from a dairy farm in Texas between February 2020 and June 2024, encompassing a total of 5,053 failed milking events out of 614,681 total milking sesssion. A generalized linear model with a binomial distribution and a logit link function was used to assess the impact of the temperature-humidity index (THI) levels, milk yield, milk speed, cow parity, and cow lactation stage variables on the odds of an FME. Similarly, a generalized linear model with a truncated negative binomial distribution and a log link function was used to evaluate the incidence rate ratio (IRR) of FME across lactation weeks and parity. Overall, teat-connection-related failed milking events were primarily due to connection time (CT; 35.1%) and increased teat-connection attempts (CA; 21.9%) . Dead milk time (DT) and automatic robot stop (AS) contributed 35.4% and 0.05% of total FME, respectively, reflecting the distribution of failed milking reasons . Overall FME rates decreased significantly over lactation weeks (IRR = 0.98, 95% CI: 0.96-0.99) as identified by IRR model. Cows with an impending mastitis event within 7 d had significantly higher odds (OR = 6.70, 95% CI: 4.59-9.75) of FME as compared with healthy cows. Longer intervals between milking significantly increased the odds of FME (OR = 1.79, 95% CI: 1.54-2.08) as compared with regular milking intervals. Milking failure risk declined across lactation in primiparous cows but increased slightly in multiparous cows, demonstrating a significant parity-dependent effect of lactation stage. This information could be used to tailor milking strategies to meet the animals' physiological requirements, and enhance robotic milking efficiency.
- Open Access
- 24 July 2026
- Published by Frontiers Media SA in Frontiers in Veterinary Science
- Vol. 13, 1856042
Evaluation of acoustic pulse and cold laser therapies for the treatment of dairy cows with subclinical mastitis under organic management conditions
Background: Mastitis remains a significant issue in dairy herds, yet antibiotic use is restricted under organic management conditions. There is a need to evaluate commercially available non-antibiotic mastitis treatment strategies. Objective: This study aimed to evaluate the effectiveness of two currently available mastitis treatment strategies: acoustic pulse therapy (APT) and cold laser therapy (CLT) on organic dairy farms. Methods: A total of 129 multiparous lactating crossbred cows with somatic cell counts (SCCs) >200,000 cells/mL were randomly assigned to one of three treatment groups: APT, CLT, and control (CON) at a commercial dairy farm under certified organic management. Cow-level milk samples were collected before milking and evaluated for SCC, fat%, protein%, lactose%, solids-not-fat%, and bacterial identification. Results from the dairy herd improvement association (DHIA) test were used to obtain individual milk production and cow removal information. The bacteriological cure rate, recovery rate, and cow survival were compared among treatment groups. Results: The CLT group had a higher bacteriological cure rate than the CON group, but only in cows with acute, contagious mastitis. Cows in the APT group tended to have lower bacterial counts than those in the CON group, regardless of inflammation status or chronic mastitis history. There was no statistically significant effect of treatment on cow recovery or survival. Conclusion: While there were some effects on bacteriological cure rate and bacterial counts in specific groups of cows, we were unable to demonstrate long-term benefits in terms of recovery rate, milk yield, or survival among cows affected by subclinical mastitis in organic settings.
- 24 July 2026
- Published by Frontiers Media SA in Frontiers in Veterinary Science
- Vol. 13, 1856042
Open AccessEvaluation of acoustic pulse and cold laser therapies for the treatment of dairy cows with subclinical mastitis under organic management conditions
Background: Mastitis remains a significant issue in dairy herds, yet antibiotic use is restricted under organic management conditions. There is a need to evaluate commercially available non-antibiotic mastitis treatment strategies. Objective: This study aimed to evaluate the effectiveness of two currently available mastitis treatment strategies: acoustic pulse therapy (APT) and cold laser therapy (CLT) on organic dairy farms. Methods: A total of 129 multiparous lactating crossbred cows with somatic cell counts (SCCs) >200,000 cells/mL were randomly assigned to one of three treatment groups: APT, CLT, and control (CON) at a commercial dairy farm under certified organic management. Cow-level milk samples were collected before milking and evaluated for SCC, fat%, protein%, lactose%, solids-not-fat%, and bacterial identification. Results from the dairy herd improvement association (DHIA) test were used to obtain individual milk production and cow removal information. The bacteriological cure rate, recovery rate, and cow survival were compared among treatment groups. Results: The CLT group had a higher bacteriological cure rate than the CON group, but only in cows with acute, contagious mastitis. Cows in the APT group tended to have lower bacterial counts than those in the CON group, regardless of inflammation status or chronic mastitis history. There was no statistically significant effect of treatment on cow recovery or survival. Conclusion: While there were some effects on bacteriological cure rate and bacterial counts in specific groups of cows, we were unable to demonstrate long-term benefits in terms of recovery rate, milk yield, or survival among cows affected by subclinical mastitis in organic settings.
- Open Access
- 8 July 2026
- Published by Springer Nature in Animal Diseases
- Vol. 6 (1) , 40
Correction: Photobiomodulation of the mammary gland for the treatment of mastitis in dairy cattle
- 8 July 2026
- Published by Springer Nature in Animal Diseases
- Vol. 6 (1) , 40
Open AccessCorrection: Photobiomodulation of the mammary gland for the treatment of mastitis in dairy cattle
- Open Access
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Investigation of poultry feces removal rate and vision-based feces weight estimation on grooved-floor panels
Efficient feces and litter management remains a critical challenge in modern poultry production systems. Traditional deep-litter systems present several challenges, including high labor requirements, hygiene concerns, and welfare issues associated with wet litter. These issues may be alleviated by using a semicircular grooved-floor design that separates birds from accumulated feces, allowing feces to be mechanically removed from the production area through robotic operation. This pilot study aimed to evaluate the daily feces removal efficiency of a robotic cleaning prototype and develop a computer-vision algorithm to estimate fecal volume for potential health monitoring applications. Grooved floor panels were placed underneath battery cages to collect fecal samples from laying hens. A mobile system equipped with rotatory brushes was developed to remove feces. A LiDAR camera was used to scan the floor surface before and after feces removal. Experimental trials showed that the mobile system removed 88.24 ± 2.37% of accumulated feces at 24-hour collection intervals. The feces removal rates were similar (P > 0.05) under different stocking densities and cleaning intervals. In addition, fecal volume showed a strong correlation with fecal weight (R2 = 0.97, MAPE = 3.04%). Both fecal volume and weight had significant differences (P < 0.05) between the different stocking densities and collection durations. These findings demonstrate the feasibility of robotic feces removal from grooved-floor surfaces and indicate that 3D-derived fecal volume may provide insights into birds’ health status such as assessing the temporospatial density in grooved floor system or estimating number of live birds in caged system.
- 6 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInvestigation of poultry feces removal rate and vision-based feces weight estimation on grooved-floor panels
Efficient feces and litter management remains a critical challenge in modern poultry production systems. Traditional deep-litter systems present several challenges, including high labor requirements, hygiene concerns, and welfare issues associated with wet litter. These issues may be alleviated by using a semicircular grooved-floor design that separates birds from accumulated feces, allowing feces to be mechanically removed from the production area through robotic operation. This pilot study aimed to evaluate the daily feces removal efficiency of a robotic cleaning prototype and develop a computer-vision algorithm to estimate fecal volume for potential health monitoring applications. Grooved floor panels were placed underneath battery cages to collect fecal samples from laying hens. A mobile system equipped with rotatory brushes was developed to remove feces. A LiDAR camera was used to scan the floor surface before and after feces removal. Experimental trials showed that the mobile system removed 88.24 ± 2.37% of accumulated feces at 24-hour collection intervals. The feces removal rates were similar (P > 0.05) under different stocking densities and cleaning intervals. In addition, fecal volume showed a strong correlation with fecal weight (R2 = 0.97, MAPE = 3.04%). Both fecal volume and weight had significant differences (P < 0.05) between the different stocking densities and collection durations. These findings demonstrate the feasibility of robotic feces removal from grooved-floor surfaces and indicate that 3D-derived fecal volume may provide insights into birds’ health status such as assessing the temporospatial density in grooved floor system or estimating number of live birds in caged system.
- Open Access
- 1 July 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (7) , 7170-7185
Dynamics of culling risk of Holstein cows in US organic dairy herds
The study aimed to analyze the culling dynamics of cows in US organic dairy herds based DHI-reported reasons. In addition, the overall cow longevity in these herds was evaluated. A total of 287,927 lactation records from cows calving between 2018 and 2022 across 434 organic dairy herds from 31 US states were evaluated. Annualized rates of live culling and death were calculated, and the distribution of culling reasons was assessed by parity, days after calving, season, year, and geographic region. The average age at culling and its association with lifetime genetic merit were also examined. The average age at culling was 62.4 mo. Reproduction was the leading reason for culling (27.0%), followed by died (13.2%), mastitis (12.7%), and low production (12.1%). Overall, the annualized culling rate was 28.4%, with 24.7% due to live culling and 3.7% associated with died. Seasonal variation was observed in culling due to reproduction, with a decline during the winter and an increase after summer. Early lactation was a high-risk period for culling reasons died, disease, and other udder problem, and late lactation was associated with elevated risk of culling due to reproduction, low production, feet and legs, injury/other, and mastitis. Cow-level factors such as calving season, parity, and cow-relative 305-d mature equivalent (305ME) milk yield were significantly associated with risk of dying within 60 d after calving. The risk of died within 60 d after calving was greater for cows calving in the fall compared with those calving in winter, spring, or summer, and higher for primiparous cows than for cows in parity ≥4. Herd-level variables, including herd-relative 305-ME milk yield and herd size, were significantly associated with the risk of dying within 60 d postpartum. These same factors, along with the calving season, were associated with the risk of live culling during the first 60 d postpartum. High-producing cows based on high cow-relative 305ME milk yield were more likely to be culled for feet and legs, reproduction, and died, whereas low-producing cows had higher odds of being culled for low production, mastitis, and udder problems. A moderate negative correlation was detected between age at culling and genetic merit indices. These findings highlight that culling dynamics in US organic dairy herds are influenced by lactation stage, parity, milk production, herd-level characteristics, and calving season.
- 1 July 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (7) , 7170-7185
Open AccessDynamics of culling risk of Holstein cows in US organic dairy herds
The study aimed to analyze the culling dynamics of cows in US organic dairy herds based DHI-reported reasons. In addition, the overall cow longevity in these herds was evaluated. A total of 287,927 lactation records from cows calving between 2018 and 2022 across 434 organic dairy herds from 31 US states were evaluated. Annualized rates of live culling and death were calculated, and the distribution of culling reasons was assessed by parity, days after calving, season, year, and geographic region. The average age at culling and its association with lifetime genetic merit were also examined. The average age at culling was 62.4 mo. Reproduction was the leading reason for culling (27.0%), followed by died (13.2%), mastitis (12.7%), and low production (12.1%). Overall, the annualized culling rate was 28.4%, with 24.7% due to live culling and 3.7% associated with died. Seasonal variation was observed in culling due to reproduction, with a decline during the winter and an increase after summer. Early lactation was a high-risk period for culling reasons died, disease, and other udder problem, and late lactation was associated with elevated risk of culling due to reproduction, low production, feet and legs, injury/other, and mastitis. Cow-level factors such as calving season, parity, and cow-relative 305-d mature equivalent (305ME) milk yield were significantly associated with risk of dying within 60 d after calving. The risk of died within 60 d after calving was greater for cows calving in the fall compared with those calving in winter, spring, or summer, and higher for primiparous cows than for cows in parity ≥4. Herd-level variables, including herd-relative 305-ME milk yield and herd size, were significantly associated with the risk of dying within 60 d postpartum. These same factors, along with the calving season, were associated with the risk of live culling during the first 60 d postpartum. High-producing cows based on high cow-relative 305ME milk yield were more likely to be culled for feet and legs, reproduction, and died, whereas low-producing cows had higher odds of being culled for low production, mastitis, and udder problems. A moderate negative correlation was detected between age at culling and genetic merit indices. These findings highlight that culling dynamics in US organic dairy herds are influenced by lactation stage, parity, milk production, herd-level characteristics, and calving season.
- Open Access
- 25 June 2026
- Published by Springer Nature in Animal Diseases
- Vol. 6 (1)
Photobiomodulation of the mammary gland for the treatment of mastitis in dairy cattle
Mastitis remains the most economically significant disease in dairy production, impacting milk yield and quality from the mammary gland and impairing overall animal welfare. Conventional treatment strategies rely heavily on antibiotics, raising concerns about antimicrobial resistance, withdrawal periods, and consumer demand for residue-free milk. Photobiomodulation (PBM), a noninvasive therapy utilizing low-level laser light, has emerged as a promising adjunct for udder health management. This review synthesizes current evidence on PBM’s physiological mechanisms, therapeutic benefits, and practical implications for dairy herd health. PBM operates through the activation of mitochondrial chromophores, primarily cytochrome c oxidase, triggering enhanced ATP synthesis, nitric oxide release, and redox signaling. These processes modulate inflammatory pathways, including NF-κB and MAPK signaling, and regulate cytokine production, thereby promoting tissue repair without broadly suppressing immune function. In cows, PBM has been shown to reduce oxidative stress, inflammatory mediators, and the somatic cell count (SCC), contributing to improved milk quality and accelerated recovery. Studies report up to a 60% reduction in subclinical mastitis cases and an enhanced recovery of 31.2% when PBM is combined with antibiotics, underscoring the role of PBM in antimicrobial stewardship. Beyond therapeutic outcomes, PBM offers welfare advantages by alleviating pain and minimizing stress during treatment. Its noninvasive nature reduces handling-related anxiety compared with conventional interventions. Emerging evidence also suggests potential benefits for milk composition (increase in de novo fatty acids) and metabolic stability, although further research is needed to confirm these effects in vivo. Despite its promise, PBM faces limitations, including variability in treatment parameters, restricted tissue penetration, and logistical challenges in large herds. Standardization of dosing protocols and cost-effectiveness analyses remain critical for widespread adoption. Thus, PBM represents a viable complementary strategy for mastitis management, aligning with industry goals to reduce antibiotic dependence while enhancing animal welfare and milk quality. Continued research into optimized applications, long-term efficacy, and economic feasibility will determine its integration into sustainable dairy health programs.
- 25 June 2026
- Published by Springer Nature in Animal Diseases
- Vol. 6 (1)
Open AccessPhotobiomodulation of the mammary gland for the treatment of mastitis in dairy cattle
Mastitis remains the most economically significant disease in dairy production, impacting milk yield and quality from the mammary gland and impairing overall animal welfare. Conventional treatment strategies rely heavily on antibiotics, raising concerns about antimicrobial resistance, withdrawal periods, and consumer demand for residue-free milk. Photobiomodulation (PBM), a noninvasive therapy utilizing low-level laser light, has emerged as a promising adjunct for udder health management. This review synthesizes current evidence on PBM’s physiological mechanisms, therapeutic benefits, and practical implications for dairy herd health. PBM operates through the activation of mitochondrial chromophores, primarily cytochrome c oxidase, triggering enhanced ATP synthesis, nitric oxide release, and redox signaling. These processes modulate inflammatory pathways, including NF-κB and MAPK signaling, and regulate cytokine production, thereby promoting tissue repair without broadly suppressing immune function. In cows, PBM has been shown to reduce oxidative stress, inflammatory mediators, and the somatic cell count (SCC), contributing to improved milk quality and accelerated recovery. Studies report up to a 60% reduction in subclinical mastitis cases and an enhanced recovery of 31.2% when PBM is combined with antibiotics, underscoring the role of PBM in antimicrobial stewardship. Beyond therapeutic outcomes, PBM offers welfare advantages by alleviating pain and minimizing stress during treatment. Its noninvasive nature reduces handling-related anxiety compared with conventional interventions. Emerging evidence also suggests potential benefits for milk composition (increase in de novo fatty acids) and metabolic stability, although further research is needed to confirm these effects in vivo. Despite its promise, PBM faces limitations, including variability in treatment parameters, restricted tissue penetration, and logistical challenges in large herds. Standardization of dosing protocols and cost-effectiveness analyses remain critical for widespread adoption. Thus, PBM represents a viable complementary strategy for mastitis management, aligning with industry goals to reduce antibiotic dependence while enhancing animal welfare and milk quality. Continued research into optimized applications, long-term efficacy, and economic feasibility will determine its integration into sustainable dairy health programs.
- Open Access
- 8 June 2026
- Published by Taylor & Francis in Veterinary Quarterly
- Vol. 46 (1) , 2685619
Evaluating machine learning algorithms based on thermal imaging and milk-based parameters to identify subclinical mastitis in dairy cows
The objective was to evaluate the potential of machine learning algorithms utilizing thermal imaging and/or milk-based parameters for the identification of subclinical mastitis(SCM) in dairy cows. Holstein dairy cows (n = 194) were enrolled, and milk samples were collected from each quarter separately (n = 776 quarter-level samples) and analyzed for milk-based parameters. Four images per cow, representing all four quarters, were captured using a handheld infrared camera. Skin temperatures were extracted using Fluke Connect software, yielding 776 images for analysis. SCM was defined at the quarter level as SCC > 200,000 cells/mL, identifying 139 infected and 637 healthy quarters. The data were analyzed to test Random Forest Classifier(RFC), Logistic Regression Classifier, AdaBoost Classifier(ABC), and Naïve Bayes Classifier algorithms. The RFC models using milk features and temperature features achieved accuracy and precision scores of 0.88 and 0.88, respectively, yielding 0.90 AUC values, with high feature importance for milk lactose and SNF. Using only thermal imaging features, the ABC model yielded accuracy and precision of 0.82 and 0.67, respectively, with 0.56 AUC. We conclude that SCM prediction using machine learning algorithms is most promising when combined with thermal images and milk-based parameters. Future studies with larger datasets and refined methods for thermal image capture and analysis are warranted to validate these findings.
- 8 June 2026
- Published by Taylor & Francis in Veterinary Quarterly
- Vol. 46 (1) , 2685619
Open AccessEvaluating machine learning algorithms based on thermal imaging and milk-based parameters to identify subclinical mastitis in dairy cows
The objective was to evaluate the potential of machine learning algorithms utilizing thermal imaging and/or milk-based parameters for the identification of subclinical mastitis(SCM) in dairy cows. Holstein dairy cows (n = 194) were enrolled, and milk samples were collected from each quarter separately (n = 776 quarter-level samples) and analyzed for milk-based parameters. Four images per cow, representing all four quarters, were captured using a handheld infrared camera. Skin temperatures were extracted using Fluke Connect software, yielding 776 images for analysis. SCM was defined at the quarter level as SCC > 200,000 cells/mL, identifying 139 infected and 637 healthy quarters. The data were analyzed to test Random Forest Classifier(RFC), Logistic Regression Classifier, AdaBoost Classifier(ABC), and Naïve Bayes Classifier algorithms. The RFC models using milk features and temperature features achieved accuracy and precision scores of 0.88 and 0.88, respectively, yielding 0.90 AUC values, with high feature importance for milk lactose and SNF. Using only thermal imaging features, the ABC model yielded accuracy and precision of 0.82 and 0.67, respectively, with 0.56 AUC. We conclude that SCM prediction using machine learning algorithms is most promising when combined with thermal images and milk-based parameters. Future studies with larger datasets and refined methods for thermal image capture and analysis are warranted to validate these findings.
- Open Access
- 1 June 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (10) , 10542-10553
Regional variation in milk production and lactation curves in response to heat stress in US dairy systems
The objective of this study was to quantify state-level differences in milk yield, milk component yields, and lactation curve characteristics across 8 major US dairy states, and to determine whether temperature-humidity index (THI) category contributed to differences in milk production. Milk production records from the Dairy Herd Information Association were obtained for California, Florida, Michigan, Minnesota, New York, Pennsylvania, Texas, and Wisconsin from January 1-December 31, 2021. Biologically reasonable records (4.54-90.72 kg/d milk yield, DIM <305 d, THI <101, fat 2-6.5%, protein 2.5-4.5%) were retained (n = 1,524,403). Daily ambient temperature and relative humidity were used to calculate THI, and test-day THI was categorized as either NO HS (THI <68), MOD HS (68 ≤ THI <80), or HIGH HS (THI ≥80). Data were summarized by state using PROC SUMMARY, and production variability was quantified using the interquartile range. State and THI category effects were evaluated using PROC GLM with significance declared at P ≤ 0.05. Lactation curves were modeled for each state × THI category using the linearized Wood's model. Parameters were estimated using PROC REG, back-transformed, and used to calculate peak milk, DIM peak, and persistency. Daily THI differed marginally among states, and the percentage of days spent in each THI category did not differ significantly by state or THI category. Milk yield differed among states; Michigan (44.1 ± 0.8 kg/d) and Wisconsin (43.0 ± 0.8 kg/d) produced the highest yields, whereas Minnesota had the lowest (34.7 ± 0.8 kg/d). Moreover, ECM followed the same pattern. Fat and protein yield also differed significantly among states, and fat yield tended to differ by THI category. Production variability differed by state, and THI category did not affect interquartile range. Lactation curve parameters differed among states for peak milk, rise rate, and decline rate. THI category showed tendencies for differences in curve rise rate and curve decline rate but did not affect predicted peak milk, DIM peak, or persistency. Regional differences in milk yield, components, variability, and lactation curve structure were observed. Overall, THI category did not alter peak milk production or persistency, highlighting the importance of regional context when evaluating milk production responses to heat load.
- 1 June 2026
- Published by American Dairy Science Association in Journal of Dairy Science
- Vol. 109 (10) , 10542-10553
Open AccessRegional variation in milk production and lactation curves in response to heat stress in US dairy systems
The objective of this study was to quantify state-level differences in milk yield, milk component yields, and lactation curve characteristics across 8 major US dairy states, and to determine whether temperature-humidity index (THI) category contributed to differences in milk production. Milk production records from the Dairy Herd Information Association were obtained for California, Florida, Michigan, Minnesota, New York, Pennsylvania, Texas, and Wisconsin from January 1-December 31, 2021. Biologically reasonable records (4.54-90.72 kg/d milk yield, DIM <305 d, THI <101, fat 2-6.5%, protein 2.5-4.5%) were retained (n = 1,524,403). Daily ambient temperature and relative humidity were used to calculate THI, and test-day THI was categorized as either NO HS (THI <68), MOD HS (68 ≤ THI <80), or HIGH HS (THI ≥80). Data were summarized by state using PROC SUMMARY, and production variability was quantified using the interquartile range. State and THI category effects were evaluated using PROC GLM with significance declared at P ≤ 0.05. Lactation curves were modeled for each state × THI category using the linearized Wood's model. Parameters were estimated using PROC REG, back-transformed, and used to calculate peak milk, DIM peak, and persistency. Daily THI differed marginally among states, and the percentage of days spent in each THI category did not differ significantly by state or THI category. Milk yield differed among states; Michigan (44.1 ± 0.8 kg/d) and Wisconsin (43.0 ± 0.8 kg/d) produced the highest yields, whereas Minnesota had the lowest (34.7 ± 0.8 kg/d). Moreover, ECM followed the same pattern. Fat and protein yield also differed significantly among states, and fat yield tended to differ by THI category. Production variability differed by state, and THI category did not affect interquartile range. Lactation curve parameters differed among states for peak milk, rise rate, and decline rate. THI category showed tendencies for differences in curve rise rate and curve decline rate but did not affect predicted peak milk, DIM peak, or persistency. Regional differences in milk yield, components, variability, and lactation curve structure were observed. Overall, THI category did not alter peak milk production or persistency, highlighting the importance of regional context when evaluating milk production responses to heat load.