Recipient Organization
VERDANT ROBOTICS, INC.
26062 EDEN LANDING RD. STE 6
HAYWARD,CA 945453712
Performing Department
(N/A)
Non Technical Summary
In the proposed effort, Verdant Robotics and Professor Aaron Smith will develop and test a novel bio-economic farm model (BEFM) to provide a mechanistic tool for modeling the economic and agroecological value created through adoption of next generation precision agriculture technologies (NG-PATs) (e.g. autonomous farm management robotics) and how the farm-level value and adoption changes as a function of policy and market variables such as farm scale, profit margins, and labor availability. The model will be tested through a comparison of field data collected from NG-PAT providers and interviews with farmers. Testing will be accomplished by data collection on farm deployments of autonomous agricultural robotics by Verdant Robotics; a typical early deployment is performed on a subset of the overall farm acreage allowing for a 1:1 comparison of outcomes on acreage utilizing NG-PATs and acreage utilizing traditional methods. Additional data will be collected via surveys of farmers.For the proposed effort, the scope of end-use study will be restricted to certain high value horticultural applications such as carrots or apples and the model is expected to be refinable and extendable to other agricultural areas such as oil and grain crops, fiber crops, and other root or tuber crops, though such an extension is beyond the scope of the current effort. The primary focus of the proposed effort will be to provide a tool to understand the effects of precision agricultural technologies at the farm and community level. Larger macroeconomic effects can potentially be modeled as population outcomes using statistical or machine learning models, such as Monte Carlo simulations, for examining the global impacts of technology policy on farm economics.Throughout history, farm labor shortages have spurred technical innovation in agriculture. In a 2019 California Farm Bureau Federation (CFBF) survey, 56% of participating farmers indicated that at some point in the previous 5 years, they had difficulty hiring the number of employees they needed for production of their main crop (CFBF 2019). This percentage was virtually unchanged from a 2017 CFBF survey. As in the past, the current labor shortage is forcing some farmers to look towards increased mechanization to maintain their productivity. Of the respondents, 56% stated they had used a labor-saving technology within the previous 5 years, with both rising labor costs and not enough workers cited as reasons by the majority of respondents (CFBF 2019).Big data use in agriculture has been shown to improve the productivity of farms. In France, SMAG has pooled 30 years of weather data history, satellite and drone images, and soil types to develop an algorithm that successfully predicts crop yields. Weed and disease identifying apps that compare photos of plants on the farm to a large database of images utilize machine learning to constantly improve the source database, which in turn helps farmers identify and treat the weeds or disease faster, ultimately increasing crop yield. Verdant's proposed technology builds on this concept by also being able to treat the weeds or disease with a robot. The robot would identify the disease or weed on its trips through the rows of crops, remember the location, and later return with a treatment targeted specifically for that disease or wood. The robot would also be able to monitor the results of the treatment over time, to determine the efficacy. The labor costs associated with the treatment would be minimal, as no person would be required to physically return to the location of the identified weed.The use of big data to improve farm productivity and ultimately, profitability, allows all farmers access to data that previously would have been limited to large, corporate farms. Given the many risks to the food supply chain, including labor shortages and climate change, increasing the number of successful small to medium sized farms serves to mitigate the risk of food shortages in the United States. Verdant's technology provides an economically viable option to small and medium sized farms looking to use AI and ML to improve their farm productivity, while reducing labor costs and offering a solution to the diminishing supply of farm laborers.Verdant's mission is to use the techniques of modern data science, deep learning, and robotics to create improved economics in agriculture and secure the agricultural production base against future disruption (also known as digital agriculture or agriculture 4.0). This is strategically important to the United States, important to the future of the American farmer, and important to food security for billions of people throughout the World.Autonomous or semi-autonomous farm robotics is a versatile and increasingly economically viable solution to both rising labor costs and farm worker shortages that will be accessible financially to farms of all sizes and that has been made possible by sweeping innovation in information technologies over the past decade. The individual plant level data gathered by the proposed robotic tool will allow farmers to target areas of need with precision, thus lowering labor costs, limiting pesticide use and increasing crop yield.Digital farming technologies like those being pioneered by Verdant have the potential to transform agricultural systems making them more productive, resilient, and sustainable. Global agriculture faces significant challenges to meet the growing demand for value-added agricultural products in a world where the total population is expected to reach nearly 11 billion by 2050. Not only does the productivity per unit land need to increase by 70% over the next 30 years, this needs to happen without further pollution of soil, water, and other agricultural/ecological systems. Digital farming addresses these problems using information technologies, robotics, and autonomy to collect and analyze data to support the most efficient farming process and to act on those analyses in real time with minimal human oversight.
Animal Health Component
50%
Research Effort Categories
Basic
50%
Applied
50%
Developmental
(N/A)
Goals / Objectives
The goal of this project is to develop and test an economic model for the adoption of next-generation precision agricultural technologies utilizing a single, high-value crop (apples) that is in the process of adopting a wide range of NG-PATs to enhance yields, lower dependence on chemical crop management tools, and to increase labor productivity in a market where labor shortages are a chronic business problem. The proposed effort will not only develop such a model in collaboration with key stakeholders, including farm owners and operators, policy makers and community leaders, and non-owner, non-executive farm workers, it will gather data from actual implementations as well as from farm acreage not adopting NG-PATsSpecifically, the model will examine the economic effects on individual farms due to adoption of different IT technologies from simple camera-based crop monitors to full or semi-autonomous crop management tools. Different farm scales will be modeled by changing inputs related to scale, which may include rule sets as well as scalar parameters such as average annual revenue, crop yield, acreage, etc.
Project Methods
Assessment of farm revenue and incomeIn order to develop a common metric that is comparable across different farm scales, crops, and locales, we are using the European frame work from Farm Accountancy Data Network (FADN) (FADN 2021), following the methods of (O'Donoghue, et al. 2016) definition of Farm Income, i.e. the "remuneration to fixed factors of production of the farm (work, land and capital) and remuneration to the entrepreneur's risks (loss/profit) in the accounting year" (European Commission Brussels) and is defined as:FI = Total output − Total intermediate consumption + Balance current subsidies & taxes - Depreciation +Balance subsidies & taxes on investment - Total external factors Total intermediate consumption represents total specific costs (including inputs produced on the holding) and overheads arising from production in the accounting year. Total external factors cover remuneration of inputs (work, land and capital) which are not the property of the holder (wages, rent and interest paid). This income does not take into account off-farm income. In keeping with the definitions in IAS 41(Deloitte 2021), income from processed farm products (e.g. cured meats, processed vegetables, plant fibers, etc) are not included in on-farm income.Farm wage thresholds Farm wages in the US are not uniformly regulated in all states. The Federal minimum wage of $7.25 is a global minimum across the US; however small farms that use less than 500 man-days of labor are exempt from both the minimum wage and overtime pay requirements. In contrast, some states have significantly higher required minimums. For example, CA SB 3 raises the farm minimum wage in California to $15/hr in 2022 and requires overtime for any workday longer than 8 out of 24 hours. Additionally, individual farms may choose to pay different rates depending on labor availability. Thus for individual states the average annual wage is estimated according to equation 1, utilizing data from the Bureau of Labor Statistics.Annual hourly wage = Total wages/payment period in hours eq (1)Farm viability Farms are divided into 3 viability classifications per the methods outlined in Hanrahan et al. (2014) - viable, sustainable and vulnerable. A farm is considered "viable" if the farm income is higher than the average agricultural wage and provides a 5% return on capital invested in non-land farm assets such as machinery and buildings. Farms are considered sustainable if they are not viable but a member of the family has off-farm employment. Vulnerable farms are neither viable or sustainable, that is, they do not produce enough profit to be viable and no other income is available. Equation 2 describes a simple calculation of the farm equivalent wage and determine viability:≥ Threshold wage eq (2)For the most part, farm viability is not expected to be a concern for base cases of carrot production. However, it is a convenient tool for identifying economic thresholds whereby a farm is non-viable in the short-term regardless of any long-term or societal value created.Other methodologies can be used as well. For example:Opportunity cost: This method tests the farmer's alternatives to working on the farm. If an hour spent working on the farm makes more money than spending an hour working off the farm, then the farm is considered viable. This method takes into account differences in opportunity individuals may experience due to regional differences in opportunity, educational differences, and so on.Cost of capital: If a farmer is able to cover the cost of capital for running the farm, then a farmer is able to, in theory, continue to indefinitely borrow to cover farm operations. In the absence of enough farm income to cover costs of capital, per eq. (2), any positive threshold income rate would exceed the farm income rate. If the farm cannot support its own cost of capital, the farm is not truly a viable enterprise.Because the focus of these models is not on determining the minimum viable size of family farms, but rather understanding threshold economics of technology adoption, farm viability is primarily used as a model threshold constraint rather than as an input on the model.Efficiency of NG-PATsIn the proposed effort, we will utilize the methods of Lampridi et al. (Lampridi, et al. 2019) to quantify the efficiency of autonomous solutions. Calculated efficiency metrics will be compared to collected data as reported by orchard operators, allowing for calibration of coefficients and further exploration of how and why theory and practice may differ. Work time equivalence: Different robotic workers are likely to perform identical tasks at different rates. Thus to compare across multiple technologies, it is important to define a normalized work efficiency. Eq. (3) provides the available time (tav ) to perform a task on a field of area A (where A is, for example, an acre).tav = p * w * t eq (3)Where p is the working period (in days), w is an empirically determined dimensionless scaling factor, and t is the number of working hours per day during which robots are available to perform the task.The functional area of daily operative capacity, C (units of acres/hr) must then meet the constraints of eq (4) for any individual robot:C≥ A*tav eq (4)Given these two equations, the number of units needed per task (n) is then defined as (where Af = the total crop acreage):n = Af*/(C*tav) eq (5)In general, field efficiency (e) of an agricultural machinery system (undependably of its conventional or robotic nature) is defined by:e = t0/top eq (6) where top is the total operation time needed to perform the task and t0 is the net operational time, namely the time needed for the machine to complete the task without any idle time under optimum conditions, per eq (7):t0 = A/v*Dw eq (7)where v (in units of rows per hour) is the operating speed of the tool and Dw (in units of rows per acre) provides the number of passes per unit area. Finally we note that tav is functionally affected by externalities intrinsic to each robot, such as charging time per acre, maintenance time, etc. Such factors are part of the scaling factor w, which may need to be determined on a tool-by-tool basis for individual crops and potentially even individual geographies or farms.Data SourcesThe United States Department of Agriculture National Agricultural Statistics Service is the gold standard of farm statistics. The US Bureau of labor statistics additionally provides state-by-state and employment type data on wages across the US. These data will be used to provide averaged economic input for the model and market prices for crops. Farm-specific data will be collected from orchards in California and Washington. Verdant has extensive customer and testing relationships with producers across the Pacific Northwest and California, and will engage the operators and owners via in-person and virtual surveys about man-hours spent on specific tasks, separating time spent on automatable tasks from time spent on tasks still requiring human labor. Additionally we will seek to identify how these labor hours are spread in calendar time. Non-replaceable tasks where the effort is fungible with respect to the calendar are considered scalable and to have a high elasticity with respect to labor availability and labor hours will be considered to have a clear 2000 man-hour per FTE equivalence. Non-replaceable tasks where the effort must be constrained in time will be considered as requiring labor resources on a shorter time frame.