Nutrient Management and Edge of Field Monitoring Conference December 1-3, 2015 Memphis, TN

1. Do Edge of Field Monitoring Results
Inform, Support, and Improve
the Predictive Models
for Water Quality Protection?
Jane Frankenberger
Agricultural & Biological Engineering
Purdue University, West Lafayette, Indiana
?

3. Do Edge of Field Monitoring Results
Inform, Support, and Improve
the Predictive Models
for Water Quality Protection?
Jane Frankenberger
Agricultural & Biological Engineering
Purdue University, West Lafayette, Indiana
?

4. Wisdom from Professor Emeritus Don Franzmeier
Field Studies
Analysis
Model Mo
del
4
Model applications
Models
Field
Study
Soil and water
conservation science
50 years ago
Soil and water
conservation science
Today
Auto-calibration
algorithms, conservation
practice effects, climate
change impacts, new
bioenergy crops

5. We use many different models
WEPP
APEX
SWAT
TOPMODEL
(or SMRM)
STEP-L or
Region 5
RZWQM

6. Landscape Unit(s)
Unsaturated
Saturated
Restricting Layer (if applicable)
Shallow Aquifer
Confining Layer
Deep Aquifer
Corn
Grass
Soil Profile
Tile Drains (if applicable)
Ditch or
Stream
What All the Models Do (more or less)

7. Landscape (Edge of
field) Processes
 Infiltration
 Tile Drainage
 Plant Growth
 etc.
Channel Processes
 Water routing
 Nutrient routing
 And others… 7
SWAT simulates processes in
“hydrologic response units” or HRUs, and
aggregates them in subbasins
HRU HRUHRU HRU
HRU HRU
SubbasinSubbasin
Watershed
Single HRU SWAT Model

8. Landscape Unit(s)
Unsaturated
Saturated
Restricting Layer (if applicable)
Shallow Aquifer
Confining Layer
Deep Aquifer
Corn
Grass
Soil Profile
Tile Drains (if applicable)
Ditch or
Stream
What All the Models Do (more or less)

9. Modeling Hydrology
Precipitation
Transpiration
(from plants)
Plant Uptake Infiltration
Percolation
Deep Percolation
Return Flow
Runoff
Runoff
Evaporation (from
leaves and bare soil)
Groundwater
Lateral Flow Tile Flow
What Models Do
Proportion of runoff vs.
infilitration determined by
Curve Number,
Green-Ampt, or saturation
Hooghoudt and
Kirkham equations

10. Phosphorus
Precipitation
Tillage
Fertilizer
or manure Filter strip
effectiveness
equations
(empirical)
Phosphorus loss with
runoff and soil erosion
Runoff interacts
with solution P in
top soil (10 mm)
Soil erosion (USLE, MUSLE, etc.
transports organic and insoluble mineral P
Plant Uptake
What Models Do
Do we have enough edge of field studies to
be confident in hundreds of process
algorithms?

11. Do we have enough edge of field
studies to be confident in hundreds of
process algorithms?
Model
algorithms
Models

12. A typical path from an edge of field
monitoring project to model paper(s)
Auto-
calibration
Climate
Change
Impacts
Effect of a new
conservation
practice
Effect of a new
crop…
Barrier 1:
Only having
access to
one data set

13. A different model:
Corn Systems Coordinated Agriculture Project

14. Central
Database
Models
Climate
Change
Impacts
Effect of a new
conservation
practice
Effect of a new
crop…
A different model: The Corn Systems
Coordinated Agricultural Project
Archived,
shared with
public after
project ends.
Additional opportunities
for data to inform and
improve models

15. This model requires agreement
on Publication Guidelines
Developed and signed by all
investigators
Primary data owners are given a
reasonable amount of time to publish
their own data first.
Secondary data users can publish
with collaboration and co-authorship
of primary data owner.

16. New project: Managing Water for
Improved Resiliency of Drained
Agricultural Landscapes
Jane Frankenberger
Agricultural & Biological Engineering
Purdue University, West Lafayette, Indiana
Purdue Iowa State Ohio State U. Minnesota
Eileen Kladivko
Laura Bowling
Bernie Engel
Linda Prokopy
Ben Reinhart
Matt Helmers
Lori Abendroth
USDA – ARS
Dan Jaynes
North Carolina
Mohamed Youssef
South Dakota
Chris Hay
Larry Brown
Brent Sohngen
North Dakota
Xinhua Jia
Jeff Strock
U. Missouri
Kelly Nelson
Supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, award
2015-68007-23193, “Managing Water for Increased Resiliency of Drained Agricultural Landscapes”,

17. The project: 2015-2020
 Goal: Agricultural producers will be able to make informed
decisions about retaining drainage water in the landscape for
increased resilience of crop production on drained land.
Project Overview

18. Collecting and synthesizing data from
existing and new research sites throughout
the Midwest and North Carolina

19. Vision: The database forms the core of
collaboration involving synthesis,
management tools, extension, education

20. Central
Database
Models
Climate
Change
Impacts
Effect of a new
conservation
practice
Effect of a new
crop…
Transforming Drainage Data:
Building on the
Corn Systems CAP database
Archived,
shared with
public after
project ends.
Decision
Support Tools

21. The Goal: Transforming Drainage
Photo from Dan Jaynes
Long-term vision:
The process of designing and implementing
agricultural drainage will be transformed to include
water retention and even water recycling.

22. Placing data in archived datasets
is becoming more common.
White House’s Office of Science and
Technology Policy has instructed major
federal funding agencies to develop plans
to make both the datasets and research
articles resulting from their grants publicly
available (OSTP, 2013).
STEWARDS, LTER, other examples…

23. Barriers to edge of field monitoring
data access cannot all be solved by
better databases.
Is on-farm monitoring data collected at
public expense a public good OR
proprietary?
If proprietary, how can data be used to
improve and inform modeling?
If results cannot be used to improve and
inform monitoring, is this an effective
public expenditure?

24. Barriers cannot all be solved by better databases.
Restrictions on USDA spatial data since
2008 is a major barrier to data sharing.
 Section 1619 was added to the 2008 Farm Bill
during the conference committee process (not in
Senate or House bill) without public hearings.
 Restricts geospatial information including field
boundaries, conservation practice locations

25. SEC. 1619. INFORMATION GATHERING.
 (a) GEOSPATIAL SYSTEMS.-The Secretary shall
ensure that all the geospatial data of the agencies of the
Department of Agriculture are portable and standardized.
 (2) PROHIBITION.-Except as provided in paragraphs
(3) and (4), the Secretary, any officer or employee of
the Department of Agriculture, or any contractor or
cooperator of the Department, shall not disclose-
 (A) information provided by an agricultural producer or
owner of agricultural land concerning the agricultural
operation, farming or conservation practices, or the land
itself, in order to participate in programs of the
Department;
or
(B) geospatial information otherwise maintained by
the Secretary about agricultural land or operations
for which information described in subparagraph (A) is
provided.

26. Is its purpose to address “privacy”?
Does not prevent access to personally-
identified farm subsidy information.

27. Is there any hope of restoring
information access?
USDA has developed means of providing
data to partners who work with USDA.
Farm Bill comes up every 5 years; it could
be changed (couldn’t it?). Changes to
legislation take a concerted, organized
effort.
How can people like us, who believe in
doing conservation better and smarter,
contribute to the effort?

28. Do Edge of Field Monitoring Results
Inform, Support, and Improve
the Predictive Models
for Water Quality Protection?
?

29. Barrier: Model complexity and
lack of best practices
 Are our models at the appropriate level of
complexity?
 ``Perfection (in design) is achieved not when
there is nothing more to add, but rather when
there is nothing more to take away.'‘
Best practices for model development include:
 Version control
 Well-structured software with comments so
users can follow the algorithms
 Even the best software has bugs. Systematic
process for addressing them is critical.

30. An aside: If Microsoft Word functioned
like most of our water quality models…
 There would be 20 different versions of Word.
– Some would work for only a few purposes.
– The meaning of certain keystrokes would completely
change between versions.
– Data (text) entered in one version couldn’t be used in
another version.
 You’d have no support when something goes
wrong because developers are focusing on writing
papers on the program.
 When the developer retired, that’s the end of the
program.

31. Solution: Modeling communities
Characteristics of a model that allow it to be
informed and improved by monitoring:
Source code available to all
Vibrant community of users and
developers
Clear path for community to contribute, as
edge-of-field monitoring studies yield new
information.

32. Software development:
Cathedral or the Bazaar
Strengths of the bazaar model
 Given a large enough beta-tester and developer
base, almost every problem will be characterized
quickly and the fix obvious to someone.
– ``Given enough eyeballs, all bugs are shallow.''
 If beta-testers are treated as a valuable resource,
they will respond by becoming your most valuable
resource.
 The next best thing to having good ideas is
recognizing good ideas from users. Sometimes
the latter is better.

33. Fully open source code should be
expected of all models used for
water quality
The “bazaar model” is preferable, because
model can never be fully validated. Many
eyes are likely to lead to better code.

34. Demand reproducible modeling studies

35. Do Edge of Field Monitoring Results
Inform, Support, and Improve
the Predictive Models
for Water Quality Protection?
?
Moving Forward

36. Agency leaders should
 Expect field data to be shared broadly (beyond
those who collected the data).
 Expect model source code to be available to all.
 Reward sharing data and contributing to
model improvement. Do not use published
papers as the primary metric for the impact of
data and model contributions.
 Support model use, or if it is outside the
agency’s mission identify another agency or
entity who can provide support.

37. Researchers should
Seek ways to use edge of field data to
inform and improve models
Engage in research and model
development communities that do this.
Avoid publishing edge of field model
applications that are not adequately
supported by monitored data.

38. Journals should:
 Provide permanent digital repositories for
supplemental information and encourage their
use for all papers.
Journal reviewers should
 Expect authors to provide all details of their
model calibration and validation, by making
supplementary information available through a
digital repository.

39. Field Studies
Analysis
Model Mo
del
4
Model applications
Models
Field
Study
Recall the two pyramids

40. Field Studies
Analysis
Model
“Tower
of
Insight”
What we need to move, in order for
edge-of-field monitoring to inform,
support and improve our models

41. Moving forward
Agency leaders
 Expect field data to
be shared broadly.
 Expect model source
code to be open.
 Reward sharing data
and contributing to
model improvement.
 Support model use,
or if it is outside the
agency’s mission
identify another
agency or entity who
can provide support.
Journals and
Reviewers
 Provide permanent
digital repositories
for supplemental
information; expect
their use.
 Expect authors to
provide all details
of their model
through
supplementary
information
available in a
digital repository.
Researchers
 Seek ways to use
edge of field data
to improve models
 Engage in model
development
communities
 Avoid publishing
model applications
that are not
adequately
supported by
monitored data.