How Precise is Precise Enough?

Alfons Weersink

Dept of Food, Agricultural & Resource Economics (FARE)

FarmSmart Agricultural Conference

Guelph, ON

January 18, 2020

Overview of Presentation

Purpose-

▪ Is the extra precision of precision agriculture worth it?

▪ Focus on variable rate application of nitrogen fertilizer

Overview

▪ Elements of Agriculture 4.0

▪ Benefits of Precision Agriculture

▪ Economics of Fertilizer Management

▪ Flat Payoff Functions

▪ Adoption Rates of Precision Agr Technologies

▪ Implications

“A fourth agricultural revolution is underway, and this

one isn’t powered by seeds or diesel. It’s all about data”

Schrumm

Previous Technological

Revolutions in Agriculture

1. Mechanical

2. Green

3. Biotech

4. Digitial (Agriculture 4.0)

Previous Technological

Revolutions in Agriculture

▪ Previous innovations tended to allow farmers to

increase the scale of operation

▪ Farmers able to increase the size of management

unit

▪ Standardized practice used across that unit

▪ (i.e. single rate of fertilizer, TMR)

Elements of Agriculture 4.0

1. Robotic advances

▪ Greater degree of automation

2. Declining costs of sensor technology

▪ Real-time measurement of factors

3. Affordable computing power

▪ Creation of decision support tools

4. Emerging Big Data analytical platforms

▪ i.e. cloud computing and machine learning algorithms

Evolution in Farm Data Collection

Evolution in Farm Data Collection

Evolution in Farm Data Collection

Evolution in Farm Data Collection

Evolution in Farm Data Collection

▪ The amount of data collected on farm through sensors

has increased dramatically over the last decade.

▪ Yield monitors, drones, smart phones

▪ The form of data has also changed from from simple

numerical values (e.g., yield) to include more qualitative

measures (e.g., type of practice, odor).

▪ Both types of variables are being gathered more

precisely and at a highly localized scale.

Benefits of Agriculture 4.0

▪ When fully realized, tools of agriculture 4.0 allow

farmers to be more precise in application of inputs

▪ Increase efficiency by tailoring inputs to the

spatial and temporal microclimates on each farm

▪ Contrast to current uniform management across

whole field or herd

But are the Benefits > than the Costs?

▪ More information has been gathered but what is

the value of the information?

▪ Data useless unless it can be curated in ways that

can be turned into useful decision support tools

▪ How does the technology improve decision making?

▪ If yes, are the benefits large enough to offset the costs

of the technology?

Process for Agriculture 4.0

1. Generate and Capture Data

2. Compute

▪ Determine appropriate treatment by zone

3. Deliver

▪ adjust treatment for observed heterogeneity

The process is only as strong as its weakest link

Crop Production Example

▪ Typical crop producer makes 40 production

decisions each year

▪ Planning

▪ Planting

▪ In-season

▪ Harvesting

▪ What is the value of additional information from

Agriculture 4.0 that improves a decision(s)?

Precision Agriculture Technologies

in Crop Example

1. Geographic

▪ Guidance systems

2. Observational

▪ Yield monitors

▪ Mapping

3. Variable rate application

▪ Seed, fertilizer, pesticide

High Adoption of Guidance Systems

▪ Ability to cover more area

in a given time period

▪ Ensures complete

coverage

▪ Reduces overlap

▪ $1/acre for tillage

▪ $8/acre for planting.

Next Step Autonomous Tractors?

High Adoption of Monitoring Tools▪ Yield monitors are now

standard equipment

▪ Maps produced at low-cost

▪ Value to identifying

chronically underproductive

areas.

▪ Identifies area of crop stress

in season

Adoption Rates Low for VRT

▪ Technical ability to apply variable rate

▪ But what rate should be applied to a given

location?▪ Should a high-yielding zone receive more or less inputs?

▪ Many pieces of information ranging from geo-

physical to climatic to production to economic

are necessary to generate a prescription for the

variable application of inputs to reap the benefits

Economics of Fertilizer Management

Economics of Fertilizer Management

What N rate maximizes profit?

0

0.5

1

1.5

2

2.5

0 10 20 30 40 50 60 70 80 90 100

Yie

ld (

ton

ne

/ha

)

Nitrogen (kg/ha)

Production function

Start with the relationship between nitrogen

applied and crop yield (production function)

Source: Pannell (2019)

Economics of Fertilizer Management

What N rate maximizes profit?

0

0.5

1

1.5

2

2.5

0 20 40 60 80 100

Yie

ld (

ton

ne

/ha

)

Nitrogen (kg/ha)

Production function

0

100

200

300

400

500

600

0 20 40 60 80 100

$/h

a

Nitrogen (kg/ha)

Revenue = Yield * Crop Price

Calculate Revenue as a function of N applied

Economics of Fertilizer Management

What N rate maximizes profit?

Subtract N cost from Revenue

0

100

200

300

400

500

600

0 10 20 30 40 50 60 70 80 90 100

$/h

a

Nitrogen (kg/ha)

Revenue

N Cost = N Price * Rate

Economics of Fertilizer Management

What N rate maximizes profit?

Payoff Function- the net return to

different rates of the input (N)

0

50

100

150

200

250

300

350

400

450

0 20 40 60 80 100

Pa

yo

ff (

$/h

a)

Nitrogen (kg/ha)

Profit-maximising

rate of nitrogen

Profit maximizing rate (50kg/ha) is associated with the

peak of the Payoff Function ($400/ha)

Economics of Fertilizer Management

▪ What influences profit maximizing rate?

▪ Output price

▪ Nitrogen price

▪ Responsiveness of yield to N

▪ Previous crop

▪ Soil fertility

▪ Weather

Does Precision Matter?

0

50

100

150

200

250

300

350

400

450

0 10 20 30 40 50 60 70 80 90 100

Pa

yo

ff (

$/h

a)

Nitrogen (kg/ha)

What range of N application rates would give payoffs

within 5% of the maximum payoff ($400/ha)?

Does Precision Matter?

What range of N application rates would give payoffs

within 5% of the maximum payoff ($400/ha)?

0

50

100

150

200

250

300

350

400

450

0 20 40 60 80 100

Pa

yo

ff (

$/h

a)

Nitrogen (kg/ha)

22 to 80 kg/ha!!Varying the N rate

within the vicinity of

the optimum doesn’t

make a big

difference to profit

Flat Payoff FunctionsThe payoff functions for many inputs,

including N fertilizer, are often very flat

Source: Nasielski and Deen

Flat Payoff Functions & VRT

0

200

400

600

800

1000

1200

0 50 100 150 200 250 300 350 400

Ne

t re

turn

($

/ha

)

N (kg/ha)

Payoff functions for 3 areas within a field (German wheat)

Flat Payoff Functions & VRT

The benefits of site-specific crop management within three

areas of a field with different payoff functions

0

200

400

600

800

1000

1200

0 50 100 150 200 250 300 350 400

Ne

t re

turn

($

/ha

)

N (kg/ha)

Implications of Flat Payoff Functions

▪ There is a large margin of error in choice of N rate

to maximize yield

▪ Note example is for a single application around planting

▪ Simple decision models are good enough

▪ Other factors will influence farmer choice of N rate

▪ Personal factors (i.e. yield max)

▪ Risk

▪ Credit constraints

Implications of Flat Payoff Functions

▪ The benefit of site-specific VRT of N is small

▪ Large adjustments in N rate to account for spatial

variability will increase revenue but not by much

▪ The technology needs to be cheap or the output

valuable

Spatial vs Temporal VariabilityWhile payoff functions tend to be flat within a year, there is a

big variation across years

Source: Nasielski and Deen

Implications of Temporal

Variability in Payoff Functions

▪ Explains why farmers tend to apply more than the

recommended rate▪ Benefit of applying more in the good years is greater than the cost

of applying too much in the bad years

▪ Value of information associated with in-season

demand for N (i.e. rainfall, temperature)

▪ Benefit from split application of N if demand for N

in-season can be determined.

Reasons for Adoption Differences

Across Precision Crop Technologies▪ GPS technologies are embodied-knowledge

technologies

▪ No additional information required to capture the value

▪ Technologies either simplify work or passively collect

data

▪ Information-intensive technologies (i.e. VRT)

requires additional skills and tools

▪ Data are inputs into decision support system and that

generates a prescription that is then applied

Adoption of Precision AgricultureAbility to derive decision rules in agriculture with

high value will be highest for:

a) Valuable output

b) Short cycle

c) Small uncertainty in production

d) High within-field or between-animal heterogeneity

e) Technologies to measure and deliver to that

heterogeneity

“In pursuing … optimal levels of decision

variables, precision is pretense and great

accuracy is absurdity.”

(Anderson, 1975)

Good decision makers ignore the bad

decisions / don’t sweat the small stuff

Summary▪ Success of precisions agriculture depends on

value created for end user

▪ not technological capability

▪ Ability to gather and compile data is becoming

easier and cheaper, but it is not enough

▪ Net value of precision agriculture depends on

the situation