University of Southern California Dissertations and Theses

Integrating GIS into farm operations at the Homer C. Thompson Research Farm in Freeville, New York

(USC Thesis Other)

Integrating GIS into farm operations at the Homer C. Thompson Research Farm in Freeville, New York

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Integrating GIS into farm operations at the Homer C. Thompson Research Farm in
Freeville, New York

by

Mary Catherine Colomaio

A Thesis Presented to the
Faculty of the USC Graduate School
University of Southern
California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
(Geographic Information Science and Technology)
December 2018

Copyright ® 2018 by Mary Catherine Colomaio

To my parents. You may not always understand what I
do, but you are always there to support me anyways.

Table of Contents
List of Figures ................................................................................................................................................................ v
List of Tables ............................................................................................................................................................... vii
Acknowledgements .....................................................................................................................................................viii
List of Abbreviations ..................................................................................................................................................... ix
Abstract.. ........................................................................................................................................................................ x
Chapter 1 Introduction .................................................................................................................................................. 1

1.1 Spatial Potential ................................................................................................................................................ 2
1.2 Local Data ........................................................................................................................................................ 5
Chapter 2 Background and Literature Review ............................................................................................................... 6
2.1 Precision Farming ............................................................................................................................................. 6
2.1.1 Geographic Challenges ........................................................................................................................... 6
2.1.2 Meeting Increasing Demands ................................................................................................................. 8
2.2 Land Sustainability ........................................................................................................................................... 9
2.2.1 Using Technology to Make Decisions .................................................................................................. 10
2.2.2 Soil Health ............................................................................................................................................ 11
2.2.3 Using LANDSAT Data in Agriculture ................................................................................................. 13
2.1 Our Use Case ................................................................................................................................................. 16
Chapter 3 Methodology .............................................................................................................................................. 17
3.1 Introduction .................................................................................................................................................... 17
3.2 Research Design ............................................................................................................................................ 17
3.3 Data Descriptions .......................................................................................................................................... 19
3.3.1Tompkins County Data ......................................................................................................................... 19
3.3.2Thompson Research Farm Data ........................................................................................................... 21
3.3.3Unmanned Aerial Vehicle Data Collection .......................................................................................... 24
3.4 Data Processing .............................................................................................................................................. 28
3.5 ArcGIS Online Web Maps ............................................................................................................................ 29
Chapter 4 Results ......................................................................................................................................................... 33
4.1 Adequate Data ................................................................................................................................................. 33
4.2 Aerial Images ................................................................................................................................................ 35
4.3 Combining Data .............................................................................................................................................. 41
4.4 Online Applications ....................................................................................................................................... 47
Chapter 5 Discussion and Conclusions .......................................................................................................................... 50
5.1 Next Steps ..................................................................................................................................................... 50
5.2 Lessons Learned ............................................................................................................................................. 54
5.2.1 Data Entry ........................................................................................................................................... 54
5.2.2 Aerial Image Processing ...................................................................................................................... 55
5.3 Future Projects ................................................................................................................................................ 57
5.4 Conclusions ................................................................................................................................................... 58
References ................................................................................................................................................................... 61
v

List of Figures

Figure 1 Homer C. Thompson Research Farm Study Area ............................................................ 4

Figure 2 Yield map (Pecze, 2001). ................................................................................................. 8

Figure 3 Final Suitability Map (Weerakoon, 2014) ....................................................................... 11

Figure 4 Geographic Analysis of Cornwall (Casalegno, 2014) ..................................................... 13

Figure 5 Vegetation water content images (Jackson, 2004) .......................................................... 15

Figure 6 Tompkins County Open GIS Data Portal ........................................................................ 21

Figure 7 Google Aerial Images ....................................................................................................... 25

Figure 8 ESRI Aerial Images .......................................................................................................... 26

Figure 9 ArcGIS Online Data Entry .............................................................................................. 32

Figure 10 Unsupervised Classification of RGB Values ................................................................ 37

Figure 11 Homer C. Thompson Research Farm Sections .............................................................. 39

Figure 12 2010 Plot Data ............................................................................................................... 45

Figure 13 2013 Plot Data ............................................................................................................... 46

Figure 14 2010 Soil pH .................................................................................................................. 52

Figure 15 2010 Soil Potassium ...................................................................................................... 53

Figure 16 Cornell Musgrave Farm ................................................................................................. 59

Figure 17 Cornell Willsboro Farm ................................................................................................. 60

vii

List of Tables
Table 1 Precision Farming in Cortland County .............................................................................. 2
Table 2 First 10 Rows of Thompson Research Farm data ........................................................... 23
Table 3 Data Relationships .......................................................................................................... 29
Table 4 Additional 10 Rows of Thompson Research Farm data ................................................. 42

viii

Acknowledgements
I am thankful to my advisor, Steve Fleming, for the direction I needed and my committee
members who gave me assistance when I needed it. I would like to thank my colleagues at
Cornell University who supported me through the final year of my degree. To the strong female
role models in GIS that helped lead the way.
ix

List of Abbreviations

AHP Analytic Hierarchy Process

CPU Computer Processing Unit

D2M Drone2Map

FAA Federal Aviation Administration

GIS Geographic information system

GSP Ground Station Pro

SSI Spatial Sciences Institute

USC University of Southern California

UAV Unmanned Aerial Vehicle

VWC Vegetation Water Content
x

Abstract

Over time, the methods and technologies by which we produce and harvest our food have
advanced. Large corporations are quick to adopt new technologies and processes, but smaller
farms can struggle to see the value in pursuing advanced technologies for farm management.
Development of a streamlined protocol for introducing geospatial technology at the individual
farm level can help prioritize operations, and help develop long-term operational plans. While
the benefits of integrating GIS software and tools are apparent to corporate farm managers, or
agricultural economists, they are not always as apparent or easily accessible to small farmers.
Using the research farm in Freeville as a case study, a developmental framework for other
farmers at Cornell University and around Tompkins County for small-scale geospatial data
integration will develop. With a focus on easy-to-obtain datasets, the procedures outlined on this
paper will articulate in a way that non-geospatial data users can understand and build upon.
This paper will examine the possible benefits of implementing GIS technology in small farming
communities of like that of the Homer C. Thompson Research Farm and discuss how it can
improve the visualization and management of small farms. While the overall impact of
introducing geospatial technology at the small farm level is not quantifiable in this paper,
understanding what data is available, and its impact on farm operations, can be beneficial in the
long-term planning and management of small farms.

1

Chapter 1 Introduction
Considering its central location within New York, and its proximity to surrounding rural
communities, Homer C. Thompson Research Farm was an ideal site for a small-scale
implementation of geospatial technology. Demonstrating to famers in the region, that GIS
technology is a worthwhile and meaningful investment. By developing these management
solutions using farm data, farmers can benefit from this change and its effects on their
operations. In a similar fashion, neighboring communities have examined farm data from a non-
geographical standpoint to provide better insight into their agricultural impact area. In a 2014
report published by the towns of Homer, Preble, Scott, and the Agriculture and Farmland
Protection Plan Steering Committee, the quality of the soil in neighboring Cortland County is
considered ideal for farming (Table 1) (Brock, et al 2014). The table uses the term “prime
farmland”, which is defined by the National Soils Survey Handbook as land that has the best
chemical and physical characteristics for producing crops. The soil quality and moisture to yield
a large number of crops, dependable water supply, and adaptability to environmental changes.
Developed with spatial understanding of the area, the farmland protection plan implements both
data analytics and geographic data. While the use of the study was to develop a planning and
environmental platform, the use of geospatial data gives the reader spatial context of the area of
focus. In a similar way, using geographic data like soil taxonomy, location of aquifers, and the
development of site specific farm data at Homer C. Thompson Research Farm can, and should,
be developed giving the farmer and researchers a greater chance for success in understanding and
implementing geospatial data practices.
Development of tools from native data, not only teaches farmers of these benefits, but it lays the
groundwork and progression into precision agriculture and high-level analytical processes
2

available with GIS. Without the fundamental understanding of geospatial technologies, the next
evolution of integration may not be beneficial and truly understood. Small-scale implantation,
like that of this project, gives context and proof of concept.

Table 1 Prime Farmland in Cortland County

1.1 Spatial Potential
Understanding how to utilize the spatial potential of farmland will help alleviate negative
environmental impacts and improve local ecology. Farmers who understand the spatial diversity
of their land will be better equipped to manage local variations in soil properties and topography.
Introducing geospatial applications with high-level precision farming concepts provides insight
in to organic and ecological data and its relationship with the farm. Attributes like weather
changes, soil health, pesticide application and tilling schedules have the ability to integrate into
day-to-day farm operations. Simple tabular data that, once compiled, can be stored at the
individual plot level, helping track operational and biological changes throughout the farm. The
research farm covers 260 acres of land, with farm data spanning back to 1962. Also available are
HOMER PREBLE SCOTT
PRIME FARMLAND 6,120 5,132 2,247
FARMLAND OF
STATEWIDE
IMPORTANCE
16,053 6, 393 7,739
TOTAL FARMLAND 22,173 11,252 9,786
3

rudimentary maps showing the development of the farm, and a secondary data store of pesticide
applications. Incorporating even a fraction of this data will give researchers, campus planners,
and farm manages spatial context to planning future operations. Figure 1 shows the study area at
Thompson Research Farm. The research farm plot sits on the edge of Freeville, NY, with much
of the surrounding area also occupied by farmland. While these operations are functional in their
present state, digitization of these records also develops a strategy for data sharing and
partnership with other branches of the university. Connections with the researchers begin during
the planning phases, but it can be difficult to provide information for past projects on specific
plots, all without compromising the data integrity of the records. Having the tools and the
support to maintain a system of control over who has access to these records builds, and develop
a secure platform for the Thompson Research Farm. For a total farm management plan, farm
managers currently rely on printed maps with post-it notes to show research plans and Microsoft
Excel to track chemical application. These maps and data are geographically sound, but provide
no accessible system of record from year to year, and no backup exists in case of a disaster or
accident resulting in destruction of the data storage. Development of an initial geospatial farming
application will kick start the process of farm data being continually stored in a location that
curates a more complete history of the property; allowing farm managers and researchers in
future years to have a cohesive idea of plans and agricultural schedules previously used on a
specific plot. The passing of institutional legacy data can also be stored in this elementary
platform. As Cornell prepares for members of their workforce to retire, the retention of
institutional knowledge is rapidly becoming an issue. If only one person fully understands how
operations are to occur, their departure creates a gap in service. By adding even a fraction of that
knowledge to this platform, processes and fundamentals can remain intact.
4

Figure 1. Homer C. Thompson Research Farm
5

1.2 Local Data
As the population continues to grow, both farmers and scientists will be more likely to
adopt new technology that will create an increased food supply on the same acreage of land,
while still preserving local ecology, and keeping costs down. Being proactive rather than reactive
will give farmers in Tompkins County the economic advantage that many of their counterparts
do not have; allowing them to pursue new opportunities for increased production of products at
their farm, through technology, new techniques in farming, or operational changes.
The movement of geospatial technology is developing a workflow for small-scale
implementation and allowing local governments to share data that they maintain with new
industries and people. Tompkins County provides geospatial services ranging from a tax parcel
locator, to providing information on the county’s agricultural land values. Knowing this data is
available for use and query may help farmers, both at Cornell and in other parts of the county,
look at their farm as more than a singular entity. Instead as a living, breathing geospatial
platform, that can continue to develop with the right tools and education. Sourcing data locally
when possible also helps develop relationship and the potential for other geospatial projects.
While geospatial data can provide a new direction for economic growth in central New
York, the widespread implementation of GIS technology for farm management has not occurred
Thompson Research Farm. Using data from Tompkins County, Cornell University, and
information provided by the research farm, this paper examines the difficulties in agricultural
data, and documentation of implementing geospatial data. Implementation workflow and
documentation will provide the research farm the opportunity to utilize the wider Cornell
University geospatial data network. This framework provides basis for any future expansion of
the project to other Cornell-owned research farms.
6

Chapter 2 Background and Literature Review

The approach of introducing GIS and GPS technology into the agricultural industry is not a new
concept. Over the last 20 years, increased use of technology in all industries has helped develop a
major shift in practice and the technological use of application in the agricultural industry. As the
use of geospatial technology in agriculture continues to grow and migrate into day-to-day
operations, agricultural groups have begun to understand the importance and value of learning
and implementing these new systems.

2.1 Precision Farming
With its large mountain ranges, natural feature, and limited room for growth, many of the
successful geospatial transitions in industries focus their attention in Europe. The unique terrain
of much of the continent is creating ideal use cases for the use of GIS and GPS technologies in
agriculture. That is not to say that precision agriculture is not happening elsewhere in the world,
but research in England and Hungary present the best research-based examples for this paper.
Precision farming is already in use in United States industrial farming, but those use cases are
large-scale operations that are far beyond the scope of this project. Most of the major agricultural
operations in the United States use tools and technology developed by manufactures with the
intended use being precision farming practices. The scope and research supplements of the
project focus on small-scale agricultural implementation of precision farming and the
development of tools for initial data integration.

2.1.1 Geographic Challenges
In Hungary, researchers worked to introduce proper GIS systems and platforms to a population
7

with little to no technology experience. In this case, the work focused on determining the best
system and process for introducing precision farming (Pecze et al, 2001); Defined by the Pecze
as ‘…a way of farming which takes into account the in-field variability, a technology where the
application-seeding, nutrient replacement, spraying, etc. has taken place to act on the local
circumstances of a given field”. Using environmental sustainability as building blocks, Pecze
developed worked to show justifications on the introduction of geospatial technology in
agricultural practices. Developing from that foundation, Pecze also states that the introduction of
GIS into these practices will help to maintain a level of sustainable practices. With elevation
challenges and limited cycling in the ecological health of plots, the use of precision farming can
help alleviate the over use of plots and farm systems. Precision control of fertilization locations
and amounts helps monitor the potential impact on the surrounding areas, while saving money
from excess use of product. Using these practices, Pecze was able to demonstrate a small-scale
theoretical application of precision farming, and its benefits. Using field data from 1999 and
2000, Pecze was able to demonstrate the use in both fertilizer application and increase in yield.
Figure two shows histograms of his results, as well as yield geography. That concept, and
understanding its value, create increased value for all those involved. While this project is similar
to the overall goal of this paper, it demonstrates the possibilities of precision farming and the
benefits farmers can expect to see, even with theoretical data. With these concepts in mind, other
attempts at having systematic planning and cultivation processes for the masses have occurred.
8

Figure 2. Yield map (RDS, 1999 and 2000), the histogram shows the distribution of the yield categories.
(Pecze et al, 2001)
2.1.2 Meeting Increasing Demands
Researchers in the Netherlands developed a geo-spatial arable field optimization service
(GAOS). The main functionality of GAOS was to develop the geometry needed to integrate into
the agricultural software; In the case of farms, this means field boundaries, and natural features.
The assisted layout pattern of fields adopted by farmers, developed over the course of a three-
year study during which farmers saw a four percent increase in income (Bruin et al, 2014).
Twenty-three of the twenty-six participants were willing to invest in a full system. A major
9

drawback to the study was the level of understanding needed by farmers to use the new software.
Like any software, time and knowledge play a factor in the usability and effectiveness of the
platform; however, with guidance, most participants became engaged in using the field
optimization service.
Both Bruin and Pecze show the application precision farming and the usefulness of the
technology in agriculture. Each displaying the continuation the agricultural and technology
implementation, but working toward a simpler process for non-uses to understand. In order for
GIS to become even more of a universal standard in agriculture, action must occur. With a focus
less on the technical application of the product, and instead, present the product in an easy-to-
understand manner. By utilizing commonly understood analogies, simple ways of displaying
data or modeling the user interface off a product that is familiar, farmers can continue to develop
a better understanding of the geospatial products.

2.2 Land Sustainability

As ecosystems change with the ebb and flow of nature, and the global population continues to
grow, food stability and agriculture efficiency are assuming a greater importance, which require
additional functionality: how do we produce more agricultural goods, like dairy and grains,
without sacrificing sustainable practices? Urbanization continues to be an ever-increasing
problem, specifically in areas with a growing population density. The conversion of agricultural
land to housing or urban landscapes challenges this question, and forces farmers to examine their
practices with a new outlook and system impact. The loss of local food sources requires the
community to outsource, putting a strain on an outside agricultural communities and increase in
energy transportation resources to bring those products to the consumer. The use of geospatial
10

technology to monitor these conditions continues to increase as the data associated becomes
more accessible to those who need it.

2.2.1 Using Technology to Make Decisions
While high-density urbanization is not an immediate threat to our study area, other
studies focusing on geospatial technology and agriculture are addressing the issue. Both
agricultural development and urbanization growth have key factors that affect long term planning
of any area. As land availability decreases, and land value increases, industries have looked to
capitalize on any available resources. Using concept developed in the 1970’s by Thomas Saaty,
the Analytic Hierarchy Process (AHP), researchers in Shri Lanka have been able to determine
well-trusted and widely used decision-making theory regarding space management (Weerakoon,
2014). Using application-based decision-making provides evidence for major projects, and help
integrate all aspects of an areas needs within the calculation. AHP’s major value is the using the
Pair-Wise comparison matrix to value judgements. This procedure is by which combined criteria
arrive at an evaluation and the same criteria allows evaluations to occur and actions taken.
Weerakoon used this platform determine the best location for specific types of farming, and
expansion of urban development from major cities. These statistical and geographic based
analyses also used natural features like rivers, lakes, and coastline as variables in the final
analysis. From these conclusions, Weerakoon was able to develop a map that demonstrates these
findings and sets the stage for future geographical analytics based on this initial mapping of the
area. Figure 3 demonstrates Weerakoons final map. Integrating a process of statistical decision
making into site-specific agriculture practices and planning can allow decisions to rank in a
hierarchy and help develop a better understanding which fields needed the most attention, and
11

taking into consideration other factors of a farm: field output, employees on staff or the overall
health of the land.

Figure 3. FINAL SUITABILITY MAP (Weerakoon, 2014)

2.2.2 Soil Health
The health of the soil is one of many overall contributing factors to the success of a farm,
as well as at Thompson Research Farm. Development of a proper soil taxonomy can aid in
determining best potential plots for research project, or areas of a farm that require additional
attention and supplemental materials. While the quantifiable variables of a soil are something
that researchers in the United Kingdom see, researchers have used soil data and geotagged social
media phots to verify the positive correlation of proximity to water sources and the decreased
amount of carbon soil (Casalegno 2014). In in the United Kingdom, inland soil’s decreased
carbon content is be directly related to the higher number of farms inland, with the decreased
number near the coast. Agricultural data developed from the United Kingdom’s ward census data
and used to calculate an overall measure of agricultural production by summing the gross
margins for all major crops/livestock. This data was displayed using £/hectare, resampled at
12

1meter resolution, and displayed at one km sample level (Casalegno, 2014). Using soil carbon
data at one km
2
resolution, and the agricultural data, Casalegno was able to develop a high-level
view of the ecologically valuable land in the UK. Figure 4 demonstrates the areas of data
collected for the study. Once this baseline data was developed, social media data provided
geographic locations to show the cultural impact on these areas, and their influence on social
media. While unconventional, this study demonstrates the human perception of plant health and
the visual impact imagery has on understanding an areas overall health. The concepts used and
the environmental relationship will create deeper dimension into this study. Viewed in
conjunction with soil taxonomy mapping, and its relation to irrigation issues, soil health is a
major area of study to consider when developing any geospatial data for agricultural application.
Since our study area is in the United States, rather than needing to produce carbon soil samples
and soil salinity tests, we can use data gathered by the National Cooperative Soil Survey. Not
developed on a singular project basis, this data relies on the collection of qualified individuals
who then report this data and samples back to the United States Department of Agriculture
(USDA). Those reports add to their data download portal for use. Soil taxonomy is a specialized
field of study, and although the data could be a variable in this project, the application of the data
at a singular location or area could warranted as a separate project to be conducted allowing
proper focus on the data to occur.
13

Figure 4. Geographical zonation of Cornwall (upper left), the distribution of agriculture, aesthetics and soil
carbon (other maps; variation scaled from 0–100), and the mean value of each ecosystem service within each
geographical zone (histograms). (Casalegno, 2014)

2.2.3 Using LANDSAT Data in agriculture
Since LANDSAT first launched in 1972, earth observational imagery has changed the
way we view aspects of the earth’s surface, and how it changes. These new technological
insights highlight the opportunities to monitor and analyze these developments. Through
multispectral imaging urbanization expansion, loss of natural land and vegetation health
14

monitoring can occur through scheduled LANDSAT data analysis. The timing and quality of
data collections can however be the major limiting factor in data collection and analysis. Clean
up of atmospheric anomalies can be difficult to process, and the proper analysis of data can be
tricky depending on the processing type. Using LANDSAT data, researchers at the University of
Wisconsin- Madison developed a workflow for analyzing the Vegetation Water Content (VWC)
in corn and soybean plantings. Based on the Normalized Vegetation Density Index (NVDI),
Jackson was able to equate the potential reflective properties of vegetation (Jackson, 2004);
However, NVDI calculations are variable and do not always produce constantly accurate results.
Using a plot of land labeled SMEX02, Jackson developed collection dates and data needs to
utilize LANDSAT data bands five and seven to develop a more accurate VWC value. Processed
data collections developed new calculations for more accurate VWC values and represent a
lower margin of error in the overall calculations. Figure 5 represents the findings from Jacksons
LANDSAT data collections.

15

Figure 5. Vegetation water content images derived from the Landsat data during SMEX02. The region is
approximately 18 by 36 km and the city of Ames, IA (Jackson, 2004)

Jackson’s research demonstrates the application of LANDSAT spectral images and determining
the health of specific areas, and developing a collection schedule based on the satellites and the
variables associated with using the data. While this project was developed to show the
16

relationship between VWC and NDVI calculations, and deriving them from the LANDSAT data,
the project can be used as a proof of concept for demonstrating the hurdles that can be
encountered when using data collected by a third party, on a differential collection schedule.

2.3 Our Use Case

Much of the success of this study relies on the understanding and cooperation of farmers to
participate in the study. Understanding the Thompson Research Farm is not only critical for this
study, but allows the basis of this papers discussion to reflect the data and natural environment at
the farm. People will not develop homes where they cannot grow food, this concept has
exceptions but the idea of connectivity and the relationships of the surrounding population is a
key element (Hart,1998). Communication between different parts of the agricultural process are
critical in developing a cohesive geographic platform and process. In a report developed by the
Towns Homer, Preble, & Scott Agriculture and Farmland Protection Plan Steering Committee,
this relationship developed into a more succinct pattern. Towns in Cortland County developed a
farmland protection plan, with the major focus being local environmental conservation, and the
understating of local zoning laws. The intended use of this study was not only to enact
environmental restrictions on certain areas of the county, but also to give local planning
originations proper background information when developing their efforts. Participation by these
communities demonstrates local support for the protection of agriculture and its economic as
well as economic impact on the area. Communities in and around the agricultural industry will
continue to learn about the impact of farms in the economic and ecological environments of the
area. These areas will develop an understanding of how geospatial technology can give them the
tools to make better decisions, and provide tools and data to garner better results from efforts.
17

Chapter 3 Methodology

Chapter 3 describes the framework of this thesis, including the data used and any processes
developed. Section 3.1 proper data implementation and the implications of data integration.
Section 3.2 describes other data used, focusing on data that needs to be collected and created for
this project. Section 3.3 explains the methods used to process the data, and examining the
transition from ArcGIS Desktop to ArcGIS Online.

3.1 Introduction

Proper implementation of a precision farming application relies on the development of data sets.
With the digitization and integration of current data records, farmers at the Thomson Research
Farm have the capability to plan long-term operations for the farm, and better advise researchers
on specific aspects of the farm. While data in this process can be simple, the visualization in
geographic context provides a new level of precision not currently in place at the farm. With no
geographic visibility, farmers must rely on institutional awareness to address issues and concerns
within the farm. With an aging employee basis, overlapping education and farm operations
needed to be stored in an understandable format. By introducing of geographic data to this
institutional knowledge transfer, data development and retention will streamline. While some of
the data is currently available through Tompkins County GIS, and Cornell University Planning
Office, much of these decisions on data were farm specific and development occurred
specifically for this paper.

3.2 Research Design
Building a small-scale geospatial, data integration does not have to be a complicated process.
18

Designs as simple as one geographic feature on a map with simple data are in essence a
geographic information system. For the integration at the Homer C. Thompson Research Farm,
the simplicity of such a system was the building blocks for various reasons. One: the users of
said system do not have any formal geospatial training. This projects intention was for non-GIS
users to have the capabilities to access geospatial data, without having to take formal training.
Two: current staffing standards and data stewardship at Cornell University does not have a
dedicated employee to develop and maintain research-based agricultural management data. As
such, consideration of designing any data integration, the longevity and maintenance of the
system need, are included in technical plans. While the goal of every geospatial professional is to
build a first-class product in any field, consideration of the maintenance beyond initial build was
a factor in system design. In his research at Cornell University, Frank Popowitch also designed a
system that, on initial build was a fully functional integration and with the proper maintenance
could live beyond the scope of his project. Using Cornell University Police as his case study, he
designed a geospatial platform that would allow for out-of-office map consumption by
Emergency First Responders, and Environmental Health & Safety staff (Popowitch, 2010).
While the project development occurred using off-the-shelf data platforms, after the tenure of the
project no steps occurred to continue building out his suggested platform. This was mainly
because no appointment of a data steward ever occurred to support the projects goals and inter-
disciplinary geospatial development. Three: the data used in this project is only a preliminary
amount of the total data available for integration in an agricultural application. Forty more years
of tabular data are available for additional digitization from the farm, and other supplemental
geographical data is available to develop a system with more depth and complexity.
Simplicity and using out-of-the-box products was used an intentional way to help promote the
19

ease of use and ease of maintenance of the system after completion. The Environmental Systems
Research Institute’s (Esri) movement in the last 10 years to server based data storage, and
widespread development of cloud based data services. Rather than developing a product that is
only consumable on ArcGIS desktop, the move to a server-based application allows internet
based geospatial data services to be included. Web based applications allow for more users, both
traditional GIS professionals and non-users, to have steady access to data, regardless if they have
a desktop version of the software installed on whatever computer they were using. ArcGIS
Online allowed for web mapping applications to embedded into other web systems, allowing
more people to use web-based maps in new ways. This move away from traditional desktop-
based systems gives new flexibility to the platform. This project developed a web based mapping
application to serve as a data hub for the Homer C. Thompson Research Farm, allowing for
better geographic visualization and relationships between year-to-year farm plot data. From the
data from the farm’s legacy storage, and simple geographic features, a small-scale geospatial
application emerges.

3.2 Data Descriptions

For this paper, the gathering of baseline information about the Thompson research farm played
a critical role. With the addition of legacy information about the farm, currently stored in paper
format, a robust geospatial platform adds to farm operations.
3.2.1 Tompkins County Data
Through the Tompkins County GIS Portal, users are able to view and select data that is of
interest to them, and download the files in a format such as .JPEG, or GeoTIFF. For this paper,
Tompkins County has provided parcels, found through the portal in shapefile format, for use in
20

any web applications developed as a part of this study. Data is available using the Tompkins
County GIS Open Data portal. Operated by the Tompkins County GIS office, this portal allowed
anyone to view or download different types of geospatial data that is publically available.
Property boundaries and addresses are among the most useful for this project. Figure 6 shows the
available data when searching the address data portal from the website. Other available data
includes transportation, natural resources and county planning data. All data in the portal is
queryable, and includes viewed using charts and graphs.
The access to local data at a simple location allows this project to exist outside the parameters
mentioned in this paper. If a farmer not part of the Cornell University agricultural system wanted
to set up a similar platform, the opportunity to gather similar data sources is available to them
through this portal. For this project, data directly from the Tompkins County GIS office is
available as a part of the data sharing between offices. These parcels not only provide spatial
context to the property, but also can identify any adjacent land owned by the Thompson
Research Farm, and be potentially utilized in the future. Long term planning and analytics will
not be a part of this project, but the potential with the data at hand is of note.

21

Figure 6. Tompkins County GIS Open Data Portal focusing on address data available in Freeville, NY

3.2.2 Thompson Research Farm Data
At the time of this paper, the Research Farm did not have any digital platform storage of their
data. Farm records from the fifty plus years of record keeping resided in filing cabinets in the
main farm building at the research facility. Sorted by year, these records were available when a
farm manager needs access to them; however, when an outside source requests data, documents
require manual scanning onto the computer. The data in the document had no dynamic
capabilities and did not represent the full capabilities of the data. Each year the data packets
restructure based on research needs, and any additional farming done by staff. In its current state,
the data did not reflective of all the research done over the farms history, and it was not possible
to develop a comprehensive history of any specific plot at the research farm. Each plot contained
22

a unique alphanumeric identifier maintained from to year, and could be used to track a specific
plots history. This value joins the tabular data to plots created from overhead images of the farm.
Collected bi-annually, Soil pH, phosphorus, and potassium values supplement the records in
correspondence of the years they are available. Fertilization type, amount used, and any
additional applications are collected, but depending on the intended use of the plot, this data
varies between parcels. General research plots may see more than one traditional application of
fertilizer depending on the research conducted, while organic plots will receive compost to help
deliver the same properties. Plots left to fallow do not always receive fertilizer applications, and
may be missing this data from the records. In addition, records show cover crop and research
crop plantings, with the latter sometimes excluded depending on the plot designation.
Additionally, the research groups tend to the plot, so there is a gap in information coming to the
farm management office. Record of any tilling or maintenance done to the plot will also reside in
these records.
Also available from the farm records are basic maps showing the general plot placement from
year to year. These maps did not have any geographic accuracy, but play a critical role in
development of an accurate geographic representation. Using UAV images, these placements
allow modifications, creating a more comprehensive geographic history of the farm. Even in the
four years of data used in this study, these maps were critical in understanding how certain plots
interacted or were completely combined, only to reemerge a year to two down the road when the
plots are separated again.
Table 3 shows an example section of the first 10 rows of the farm data after entering plot
information into Microsoft Excel.

23

Table 2. Excerpt from plot records after they have be added to Excel
PLOT_
ID
PROJECT CROPS_GROWN TOTAL_
AREA
SQ_FT ACRES PH P K
A1 NEWSS Multi Crop for
Herbicide ID
38,850 38850 .89 5.7 16 175
A2 M.MAZOUREK MISC SQUASH 23,600 23600 .54 6.7 31 385
N1A M.MAZOUREK CUCURBITS
WATERMELON
20640 20640 .47 5.9 31 250
N1 M.MAZOUREK MISC CURCUBITS 74000 74000 1.7 6.3 37 2880
N2 R.
BELLMDERS
BEANS SNAP AND
DRY PEAS
75420 75420 1.73 6.2 39 360
N3 R.
BELLMDERS
POTATO 73260 73260 1.68 6.2 38 335
N4 BELLMDERS/
GALLOW
N/A STONEPICKED N/A 70920 1.63 6.4 49 410
N5 BELLMDERS N/A N/A 74000 1.7 6.0 41 390
N6 BELLMDERS WINTER SQUASH
PUMPKINS
N/A 102000 2.34 6.4 50 405
N7 R.
BELLMDERS
SWEETCORN 5.
200 FT
56000 121250 2.78 6.6 40 355
N7N FALLOW CLOVER N/A 63560 1.47 6.7 30 290
N8 FALLOW RED CLOVER N/A 86520 1.99 6.6 6 175
N9 A.
RANGARAJON
SWEETCORN N/A 150000 3.44 6.6 33 280
N10 FARM RYE CLOVER N/A 61800 1.42 6.8 8 2010
S1 O. HOEKENG TOMATO
CUCURBITS
CRUCIFERS
SNAPBEANS
43600 43600 1.0 6.7 5.2 235
S2 FARM/ANN R. ESTABLISH COVER
FOR 2011 SEASON
N/A 45000 1.03 6.9 65 255
S3 K. PERRY POTATO 47200 47200 1.09 6.2 45 355
S4 R.
BELLMDERS
STRAWBERRIES
CRUCIFERS
N/A 56200 1.29 5.7 40 410
S5 R.
BELLMDERS
CAB. EGGPLANT
PEPPERS
TOMATOES
50968 50968 1.17 6.2 44 340

For this paper four years of the data, 2010 – 2013, were migrated to Excel and integrated as part
of this project. This interval determines the best length of time needed to digitize the data, as
well as understand its usefulness.
24

3.3.3 Unmanned Aerial Vehicle Data Collection
For an additional layer of detail and visualization, this project used an Unmanned Aerial Vehicle
(UAV). Cornell University Facilities and Campus Services (FCS) owns and operates a DJI
Inspire 2 UAV, and has worked with the Thompson Research Farm in the past, as a facility for
UAV flight practice and data collection. For this project, FCS allowed the use of the UAV.
Using the UAV, this project will collect overhead images to process into topology, three cm
accurate GeoTIFF files, and aid in the digitization of individual farm parcels. While not a critical
aspect of this project, overhead aerial images in Google Maps, Esri, were not adequate for
determining plot edges. Figures 7 and 8 represent images in Google Maps and Esri respectively.
The lack of clarity in these images did not give enough spatial context to be useful. Beyond the
initial data integration, these 3
rd
part images are not useful in any additional analytics. RGB data
is capable of multi spectral analytics but it not always the intention.
25

Figure 7. Google Aerial Image quality
26

Figure 8. Esri Aerial Image Quality
27

Scheduled image collection was determined on weather allowances, with batches of data
collected on the same day to introduce a degree of consistency between the images. The first
section of images dates July 2018 were a proof of concept for the data collection timeline. The
remaining collection occurred over two days in August 2018. Weather patterns were the major
contributing factor to data collection schedules. Quadcopter flights can occur in a variety of
weather patterns. Aerodynamics of quadcopter flight favor stable heavy air mass, allowing the
UAV to move easily in the sky, and the pilot to have better control over the device. These stable
air patterns have a high humidity content, giving the quadcopter propellers better lift in the air,
with low wind from the air mass being slow moving. Overcast days with the cloud cover
lowering reflections from water, plants and building in any images collected. Overhead images
taken on a sunny day will still be useable, but there is a risk of distortion from shadows,
especially around buildings and any large objects. This will be especially obvious after
processing, and proper match points fail to generate. Using Drone2Map (D2M), an ArcGIS
companion software, overlapping images generated new compiled images. The D2M platform
uses geo-located images from UAV cameras to develop 2D and 3D products. The user has the
ability to select specific parameters for a data processing query, based on their needs and the
speed of the processing. For this project, the parameters selected that allowed singular
orthomosaic file generation, while simultaneously processing topographic data files. These
processing options can be detailed in their nature, or as basic as selecting a checkbox before
processing. This ‘behind the curtain’ processing that Drone2Map provides allows for a dedicated
software to only process aerial imagery, while not requiring manual match point selection by the
user. Figure 10 shows the processing options when setting up an orthomosaic project. Once
28

completed, the orthomosaic images served as a basis for new plot boundary data, generated form
the yearly plot maps. The orthomosaic verified these locations as well as provided any visual
clues to plots modifications or combinations in years past. These plots were then assigned plot
identification based on the alphanumeric tags used in the plot records. Plots then have the ability
to join with the tabular data and brought online.

3.4 Data Processing

Data required to develop a program management system fell into three phases: Foundational,
Data Creation, and Data Compilation. Foundational data consisted of files useable in their
original format and currently accessible by anyone with an internet connection. Data Creation
and Compilation focused on the data needs for this specific project at the Thomson Research
Farm. Creation of data focused on collectable datasets like aerial images and topography from
UAV technology, and field boundaries gathered from the aerial excursion. Data compilation is
the digitization of historical planning records that are currently in use at the research farm. This
will allow the planning of farm operations to consider historical elements. This data was stored
on a secure ArcGIS Enterprise Server developed by the Cornell University Planning Office for
use by units across campus. Chart 3 demonstrates the relationship between Cornell’s server data
storage, and the link to ArcGIS Online. Using all of these data sources, an ArcGIS Online web
application can give the farmers insight and a geographic visual of plans.

29

Table 3. Cornell’s Server relationship and its connection to ArcGIS Online.

3.5 ArcGIS Online Web Map

Developed as the next online iteration of ArcGIS Desktop, ArcGIS Online provided a platform
for geospatial data to develop into new online maps or mapping applications that look and feel
like a traditional webpage. Not to be confused with ArcGIS Portal, ArcGIS Online provided a
secure web environment for groups to store, manage their online mapping, and associated
applications. For this project, Cornell University hosted the final location of the server side data,
as well as hosted the online locations for the application build-out. Using Cornell University’s
30

preexisting accounts helped promote the longevity of this project after its completion, as well as
remove the need to migrate the data off the USC ArcGIS platforms after this project was
finished.
After data compilation from UAV images and farm documents, and subsequently developed into
each study years’ respective maps, datasets were prepared for online processing. Data uploaded
to the web must follow specific parameters set by ArcMap. The user runs the risk of the data not
processing to the web if these specific parameters are not completed. All data uploaded to the
web must have basic metadata, allowing for better tracking and management of data files once
uploaded to the web. It is possible to bypass some of these processing steps, but it may produce
errors with data down the line. Using the Cornell ArcGIS Server as an administrator, all data
aspects of this project reside at the server level as package files, keeping the symbology and
relationship between data maintained. The advantage to using ArcGIS server, rather than
uploading the data directly to the web provides an added layer of security to this project. Besides
having server based authentication, data on the server has the ability to integrate into ArcGIS
Portal. Similar to ArcGIS Online, ArcGIS Portal allows the Cornell Central IT group to set up
login parameters based on Cornell University Net ID accounts, restricting access only to those
who receive permission. At the completion of this project, those are not necessary steps, but
designing the system with such flexibility will allow for the option later on if it as needed.
Once verification occurred and uploaded to the server was complete, data was then available
from ArcGIS Online. It is possible to add data directly to the map and modify based on your
needs, but for this project, the REST location of the data within the server provided the best
connection. Figure 9 shows the selection of adding data from the server in the ArcGIS Online
environment. Adding data this way, kept the processing load of the online map down, and allows
31

for faster maps with higher data content. This was especially helpful with the aerial images,
adding a significant processing load to any maps they reside in. With the map data added, and
maps for each year compiled, the web mapping application was then developed. ArcGIS Online
provided many choices for web mapping applications based on the need of the project and the
data. For this project, the Story Map Basic provided the proper environment to develop a tabular
web mapping application. These map designs allowed for multiple maps and data sets at the
same location, emulating a webpage developed in a traditional platform. Each tabbed section
represents the years included in this study, and populated with the appropriate web map. Setting
up a tabular platform also allowed the project room to develop and grow as it continues, with
minimal effort and the same web address. Each web map is queryable and can display the tabular
data joined from the plot identification files in the attribute section. Look and feel, controls (or
widgets) were customizable through a development wizard that requires no programming skills,
and allows for real time updating of the application. Once completed the application has have a
stable web address, and can be embedded into any website for further use across the web.
Additional data easily merges into the platform as needed, developing a central hub for any data
or analytics preformed on the project, or future endeavors.

32

Figure 9. Adding data to ArcGIS Online from a REST service

33

Chapter 4 Results

When beginning the research for this project, a major concern of the project was the
procurement and adequate data to develop a geospatial database for agricultural data. During a
previous iteration of this project, attempted was made to develop a countywide agricultural
database for Cortland County. With data provided by the county, a preliminary data listing was
established; however, without specific agricultural information, the database was not functional
for in its intended use. The project ultimately was not successful due to the lack of data
resources, and the vast scope of the project.
4.1 Adequate data
During a practice flight with the Cornell Facilities UAV team, the farm managers at the Homer
C. Thompson Research farm discussed the opportunities that UAV technology would bring to an
agricultural operation, especially one tied to research projects. At these initial discussions, the
farm managers acknowledged that the farm had been keeping records since the 1960’s with all of
the data following similar standards. Each plot assignment generated with a unique alphanumeric
value, maintained from year to year. In the case of multiple plots merging, or disintegrating, new
ID names are added or removed as needed and recorded in that year’s data packet. Also recorded
in each plot’s record is soil health, crops, cover crop, and any notes pertaining to the
maintenance or history of that growth season. Consistency of information was a key factor when
examining the plot data. Plow dates and research crop listings did not always have complete
records, and some information was complicated to decipher. That is not to say the data itself was
complicated, but the handwritten notes sometimes became crammed or jumbled based on the
authors handwriting. Deciphering these data sets required time and patience to ensure each notes
formatting matched to create a cohesive tabular dataset. Each tab in the file is represented a year,
34

to keep the records separate and viewable by year, but also to create a cohesive map when joined
to the plot polygons later on.
In the initial phases of this project, the intention was to integrate as much data as possible. With
over fifty years of data to select from, this quickly became a non-viable angle for this project.
After reaching this first conclusion, a brief consideration of working with ten years of data came
up. At the time, the feeling was that this would be a summative amount of history, and display
the conceptual aspect of this project.
Involvement of the mangers of the Thompson Research Farm was an essential part of this
project. Receiving insight on farm operations, answering questions about records, and providing
history for the farm not contained in the documentation were all beneficial to the project. One of
the major insights garnered from conversations is the amount of data that was actually contained
within each plot document. When beginning this project, the project understanding was that data
input would be a major area of work. The lift to convert the entire farm to a digital platform
would take beyond this projects scope and time limitations, but most alarming was level of detail
that was contained within the records. During the data management phase of this project, the
decision to use only four years of records became the best option; this adequately displays the
functions of the system, and provides a benchmark timeline for work that the rest of the records
can potentially take.
Each plot record contained the following data: plow dates, fertilization treatments, crop and
cover crop plantings, and any miscellaneous operations that took place during the year. Many of
the plots contained the minimum data fields, providing enough to fill in each of the fields, but
not diving into depth on that year’s operations. Other plots contained little to no information. The
plot grew just fallow that year, or did not receive much attention. Fallow refers to the practice of
35

tilling a plot of land, but leaving it unsown in order to allow the soil to restore its fertility. In the
case of the Thompson Research Farm, this was typically clover or rye left to grow to its own
accord. Plots with the most information were difficult to input as the large amounts of records
would not properly display in pop-up views later on in the project, and when viewing the data in
an excel document. Much of this additional information was recommendations on planting
settings, comments about farm employees or volunteers that were involved, and while relevant to
the farm history, is not applicable to the application of this project. These ‘non-essential’ items
became omissions from the data input, as their presence was inconsistent and not critical to
understanding the rest of the plot record. The miscellaneous history of the plot did contain
comments on the health and history of the plot. These included information on flooding, changes
to crop planting, or notes on tilling and harvest schedules.
Also included in each packet of plot data, are maps designed from Google images and tools in
Adobe Acrobat. Serving as inspiration and a basis for plot placement, these rudimentary maps
provided a benchmark for this project. Determining the usefulness of paper maps and documents
in their original state, and using the basis of precision farming fundamentals, these records act as
the foundation of the farms history, and as such played a pivotal role in this project.

4.2 Aerial Images

The use of UAV technology is becoming increasingly common in agricultural practices. The
ability to access the center of a densely planted field to examine plant growth without disturbing
the outlying rows. LANDSAT data from the USGS, near infrared cameras and overhead images
all can play a role in examining the overall health of a plot. Using LADNSAT bands two or
three, farmer could run an unsupervised or supervised classification with band two data to
determine specific amounts of vegetation or examine chlorophyll production to monitor the
36

health of plants with band three. LANDSAT data schedule, determined by the USGS, focuses
on a path delineated on the satellites orbit around the sun. Data collection is frequent, but there
is no control over when the data schedules; therefore, it is possible for every images collected
during a planting season to contain cloud cover that does not allow for the analysis of
vegetation.
For the development of a geospatial platform at the Thomson Research Farm, only technology
allowing overhead RGB images was available to use data collection. While this data is able to
utilize similar data component, the result is not as accurate as a LANDSAT dataset. Figure 10
shows an RGB image, classified in an unsupervised format. Cornell University Facility and
Campus Services owns and operates the Inspire 2 quadcopter, capable of thirty minutes of flight
time, four hundred foot data capture and up to 4k video quality. During initial planning
discussions of this project, farm managers that up to date aerial images would provide up to date
field boundaries, show the growth of the farm, and help better understand the history seen in the
visual aspects of the landscape.
37

Figure 10. Unsupervised Classification of an RGB image.
38

To prepare for the initial data collection, the project explored the option of automating the UAV
into order to collect consistent images, and develop a standard of flight specifics, repeatable as
needed. Because the entire farm is two hundred and sixty acres, flying to take the images
occurred in phases. The farm’s geography and naming conventions allowed the creations of five
sections: Certified Organic, Freeville Plots Farm, Entomology Plots, and Terwilliger Plots.
Figure 11 displays these sections in relation to the total farm area. Collected in July 2018, the
first set of images of East Farm are the partial result of testing maneuvers with the Cornell UAV
team. Using the DJI Ground Station Pro (GSP) App, a six-acre test pass at three hundred eighty
seven feet generating a resolution of three cm images became the optimal flight settings.
Generating seventy-four images, this first pass was a success. By separating the automated
flights, any technical difficulties were resolved, and farm section collection at a sufficient rate.
39

Figure 11. Thompson Research Farm Sections
40

One of the difficulties of UAV flight planning is determining the weather at the time of data
collection. Like airplanes, UAV have ideal conditions and physical limits in the air. In ideal
conditions, low wind speed, high humidity, and partly cloudy skies produce the best
circumstances for flight, and overhead images. The cloud cover reduces glair on the overhead
images, producing images with flat surfaces, and low shadows. These low shadow images are
best for digitizing data from images, as it is possible to take photos on a sunnier day, but images
taken are only ideal for ‘glamor shots’. Meaning, ones used for promotional material for a
location, rather than analysis or record. During the mission planning, date selection occurred to
generate a workflow, but with the unpredictability of weather, any day selected might not be the
best candidates for data collection. The Federal Aviation Administration (FAA) acts as not only
the authoritative power for commercial and hobbyist UAV pilots, but also provides up to date
weather and barometric readings for local airports. These readings helped arrive on a decision
before driving to the flight location, if conditions were ideal enough to consider flying. Under
the Part 107 regulation, this project was responsible for the flight planning, aircraft safety, and
proper support staff for any flight. The project also retained the responsibility to check the FAA
for weather warnings and any notices handed out by region.
After successful data collection, imagery processing began using ArcGIS Drone2Map (D2M).
D2M allows the georeferenced images to be matched based on location and create a cohesive
ortho image. Processing for the flights happened separately, again to allow for any technical
difficulties. Once completed and quality checked for continuity, combined images form a single
file. Having this background image allows for creation of accurate field polygons to be and
analysis of farm management.
41

4.3 Combining Data and Maps

Besides gathering and maintain plot records, each year of records contains a simple map of the
farm to provide a visual aid for navigation and planning. These maps range from hand drawn
sketches, to documents developed in Adobe Acrobat with hand written notes. While effective
for short term planning, these maps provided context and history to the farm, as well as the
impact of certain research projects from year to year. As a part of this project, digitized plot
records migrated to Excel, and it was necessary to bring in the plotting structure as well. Chart 4
represents a section of this data as shown in Microsoft Excel. From year to year, plot structure is
subject to change; whether the change occurs due to a natural occurrence that forced the farm to
shift, increased research demands, or the planned rotation of plot structure. After completion of
the overhead flights and reconciliation of the imagery collected, polygon development occurred.
First, using these new images, the polygons boundaries were developed to reflect the current
state of the plot structure. Within the feature class, fields are added for plot ID and usage for the
year included in the study. When completed, these polygons will represent a high-level view of
the farms history, but did not have to contain any of the plot specific records. Record
delineation from year to year became a part of the red with a simple Y/N variable, to reduce the
amount of data stored within the file.

42

Table 4. Excerpt from the plot records from Microsoft Excel

PLOT_I
D
PROJECT CROPS_GRO
WN
TOTAL_AR
EA
SQ_F
T
ACRE
S
PH P K
S6 J. CIOVANNONI TOMATOS N/A 43125 .99 6.3 38 21
5
S6A FALLOW N/A N/A 1750 .04 3.6 25 28
5
S7 G.
BERGSTRON/CHRIST
INE LAYTON
SWITCHGRAS
S
N/A 24970 .57 6.4 63 80
S8 FALLOW N/A N/A 29140 .67 6.3 31 25
0
S9 M.MAZOUREK/M.
GLOS
CUCKS
SQUASH
N/A 9250 .21 6.0 39 28
5
S10 BELLINDER MUSTARDS
COVER CROP
N/A 29400 .67 5.7 38 31
5
S11 M. MAZOUREK WATERMELO
N
N/A 31490 .72 6.3 48 33
5
S12 FARM SWEETCORN N/A 44000 1.01 6.3 31 16
0
S13 FALLOW N/A N/A 41760 .96 6.6 53 12
5
S14 FALLOW RYE CLOVER N/A 67500 1.55 6.5 40 13
5
S14A FARM CUCURBITS
PUMPKINS
N/A 22750 .52 6.5 40 13
5
S15 FALLOW N/A N/A 66600 1.53 6.3 32 21
5
S16 M. HOFFMAN/J.
GARDNER
CURCUBITS 10850 10850 .25 6.4 26 80
S17 NEWSS MISC. W/
GRASS
SEEDING
N/A N/A 6 N/
A
N/
A
N/
A
S18 NEWSS MISC. W/
GRASS
SEEDING
N/A N/A 6 N/
A
N/
A
N/
A
S19 NEWSS MISC. W/
GRASS
SEEDING
N/A N/A 6 N/
A
N/
A
N/
A
S20 NEWSS MISC. W/
GRASS
SEEDING
N/A N/A 6 N/
A
N/
A
N/
A
S21 R. BELLMDERS BASIL
CARROTS
SPINACH
RADISH
N/A 45600 1.04 7.0 80 95

43

Once the existing polygons resided in ArcMap, and before assigning Plot ID values, the project
returned to the original maps in the plot documents. The suggestion came to the project to scan
the original map documents, georeference them based on the overhead images, and use them to
complete any maps. While this could work in theory, the original maps documents are
extremely simple, with some containing little to no geographic reference points. These original
maps do however provide context to plot naming conventions and help in verifying the
locations and total number of plots year-to-year. Using these documents, the project verified
plot location based off the new ortho images, and added the proper plot id. Before joining to the
digitized plot data, the naming conventions between plots and plot records needed to be
verified. This verification ensured no plot lacked data, and helped clean up the data to better
match is geographic counterparts.
Once the polygons received a proper plot identification, and the location was verified using the
original documentations, the file was finally ready to join to the new excel file. To prepare for
upload to ArcGIS Online, each year of data displayed independently within an ArcMap
document. Figures 12 and 13 show the geographic differences between data used in 2010,
versus data used in 2013. First added to the map document, was the polygon feature class.
Using the ‘joins and relates’ feature, each year of the Excel sheet was joined to the polygons
layer dependent on what year the map was being built for. Because of the possibility of there
being more data in the yearly packet, than plots on the farm, the data set was verified before the
join took place, and the records were kept intact; meaning, if a plot id in the excel sheet didn’t
match to a specific polygon, the information still remains the attribute table of the polygon
layer. Once the data verified, and joined to the excel document, the polygons layer was
symbolized based on that year’s usage. Using the usage category stored natively in the plot
44

polygons, data symbolization allowed only plots used in that year to appear, giving a brief high-
level overview of the farm for that specific year. Using the ‘Used_YYYY’ field, the symbology
of the map can determined simply using the Y and N characters in the field. Once properly
symbolized by year usage, the data is ready to upload to ArcGIS Online. It is possible to
develop a symbology based on usage, or specific crops used in a year, but for the baseline
implementation of this project, a simple usage map will meet the needs of the project.

45

Figure 12. 2010 Thompson Farm Plots
46

Figure 13. 2013 Thompson Farm Plots
47

4.4 Online Application

Development of proper data and geographic features are a major part of this project, and if the
users fell in to the traditional ‘power user’ group of ArcGIS users, then maybe the project could
have potentially stopped here. However, users were not fully equipped with the knowledge of
GIS, but rather have simple technology skills that allow them to access the needed data in a
geospatial platform. The existence of applications such as ArcGIS Online, ArcGIS Portal, and
ArcGIS Collector do just that. For the purposes of this project, ArcGIS Online allowed for
public access to the maps and data published as part of the project. In addition, Cornell
University staff will have the accessibility to update and maintain the data after the completion
of this project.
Because Cornell used both ArcGIS Online and ArcGIS Portal in their data services, these
applications would best meet the needs of the end users, Thompson Research Farm and its
researchers. ArcGIS Portal configuration accepts only users cleared by Cornell staff, and have
proper privileges through Duo Two-Factor Authentication. The University central IT staff
configure this login system and as such, the process to gain access to the data is more
complicated. Access to ArcGIS Online is a more streamlined process. Having the correct login
and password for specific account, users are able to access data, create new maps, and develop
web applications. The other approach from ArcGIS Online is the use of public maps. Maps
without the need for any login or security.
At the start of this project farm managers discussed access and security of this data. Who should
have the ability to access these maps and data? For the purposes of this project, and the limited
amount of data used, it was determined to build out the initial proof of concept in a public
environment. This way all players have access to the information and can have input on the
48

final design of the web application. After the completion of the project, if the project is
continued, there is a possibility of moving to ArcGIS Portal to have stricter access to the full
breadth of data available.
With the data fully prepared, and the platform chosen, datasets could upload to the web. Starting
from a map document for each specific year, the polygons layer and the aerial imagery could
upload to the Cornell University ArcGIS Online account. Not only did this allow data
management to carry out, but it also gives me access from an administrative account, and
associates the data with the university. Initially, the data resides in the Cornell University
ArcGIS Server to retain security standards. When the web map development begins, the REST
link to this feature can connect the two applications. This workflow also allowed data utilization
in the Cornell University ArcGIS Portal when applicable, keeping the data links open for other
analysis. Publishing the data required the data service to have a name, a brief summary of the
data to be included, as well as at least one descriptive tag. Adding each of these ensures the data
had some metadata attached, and is queryable in ArcGIS Online. Analyzing the data was
required before it its published allows for the correction of any errors. The most common are
those previously mentioned, as well as the addition of feature templates. Templates are a simple
remedy of starting and stopping and edit session within the map document, and reanalyzing the
data.
Once published to the ArcGIS Online platform, a data service could populate a web map. Once
added to the web map, any edits to the symbology occur. For this project, four separate web
maps for each year of data examined were developed. Sharing settings of any new map received
an assignment to full public access for the duration of the project. Using the 2010 data as a
starting template, the web map became the basis for the web application. Using the Story Map
49

Series template, the web map emulates a similar style to the excel sheet. Each year could be
contained within its own tabular section, with easy navigation between panes. The web maps
created for the different years of focus become their own tabs within the application, simply by
adding them to the section when prompted by the site. Configurable application tools were an
available added to allow additional capabilities. Printing, measurement, query, and location
assistant were all applicable add-ons to this project, and others were available under the settings
toolbar. Customization of these tools gave the user a streamlined experience. Once the web app
contained the web maps, and the additional features configured, the application could save and
published for viewing. Because the ArcGIS Online environment could be considered a ‘what
you see is what you get’ development environment, the final version of the web mapping
application functioned as a standalone site, or can be easily embedded into a site. This is in
contrast to a full-blown write up in a programming application. That is not to say that it was
impossible with the tools provided be Esri, but for the simplicity of this application, using a tool
that provides the necessary links internally was beneficial. Once the web mapping application
was live to the public, the first phase of integration of small scale GIS at the Thompson
Research Farm was complete. The final web maps location is at http://bit.ly/2E9Ntrk .

50

Chapter 5 Discussion and Conclusions
Bringing the Thompson Research Farm to a preliminary geospatial data integration acted as the
first phase of full geospatial data integration at the farm. Because this study only covered a
small portion of the data available for integration with the plot polygons, the entire project of
geospatial data integration can continue. Adding any number of the remaining years to the
database will continue to bring geographical context to the history of the farm, and develop a
wider breadth of data available for analysis.

5.1 Next Steps
Based on the time spent on this project, digitizing four years of data takes roughly four weeks of
steady work, if the individual works at least three hours a day on the data input. If a crew of
student workers or other staff received the task of digitizing data, it is not to say that the
remainder could migrate to a digital platform within a month’s time. This allows for steady data
input, quality control, and verification of naming conventions within each year’s data packets. A
suggestion by a colleague at Cornell that scanning the accompanying paper documents may be a
potential solution for data input time. Rather than substitute all data entry with these documents,
specific plot scans would attach to the polygons. This feature within ArcMap products allows
non-geographical and tabular data to be associated and stored within the same database.
Scanning each plot record and associating the data would take more time, and each year’s files
would have to be distinguished when adding them to a data association. Because the feature
allows for many data types to be stored, images could also be a potential addition to the
geographic dataset. Images of specific areas of the farm may provide context to some of the data
contained in the set, but if the intention is to focus on the entire history of the farm, this may
51

quickly become too much data, with too wide of a chronological period to be accurately
contained in ArcMap. Development of a file based photo structure may be of more use, and
allow project specific images to remain together.
One area of data that is included as apiece of the initial data to expand upon is the addition of
soil data to any geographic analysis. Contained in each plots records are the phosphorus,
potassium and pH of the plot, but no specific data is included as to where the collection took
place. Figures 14 and 15 show soil data for pH and potassium levels in 2010. This data
collection occurs on a biennial basis. Development of a more comprehensive view into the soil
taxonomy of the farm is a possible data creation. The USDA has developed soil polygons to
represent the soil taxonomy across the United States, but these studies only update with
submission of new findings. Rather than relying on low-level soil data, the research farm could
embark on a full-scale soil identification project, developing its own soil polygons based on
their data and the USDA data. Such a project would add continued depth to the farm, as well as
develop a granular history of the farm, and show the relationship between certain research
projects and their impact on the surrounding soil.
52

Figure 14. 2010 Soil pH
53

Figure 15. 2010 Soil Potassium Levels
54

Supervised and unsupervised classification of aerial images area additional areas of research as
this project progresses forward. While the traditional idea of image classification is synonymous
with LANDSAT data, it is possible to establish a classification grid though RBG spectral
images taken from everyday cameras. Using the ‘Iso Cluster Unsupervised Classification’ tool
in the Spatial Analyst extension, ortho images can be classified based on pixel value, generating
a total sum of specific pixel values once calculated. These classifications and calculations allow
researchers to see the estimated total of vegetation in a given area. This addition to research on
plant growth charts, and giving granular identification of incremental growth if a schedule of
image collection can be set. While this is possible to do in single plots, the wide scale
classification also allows for large sections of farm, or whole farm, classification to process with
a single data analysis.

5.2 Lessons Learned

5.2.1 Data Entry
While a simple project, this paper still laden with hurdles. Initially, this project set out to be a
full-scale precision farming program. Helping farmers understand the benefits, cost savings, and
technological advances of the move to precision farming, and developing a platform for the
move was the fundamental outcomes during the projects infancy. Quickly that became a non-
realistic achievement during the projects parameters. Determining how to scale this project
down to an integration rather than a migration seemed to be the only logical path. Reaching the
conclusion that with too large of a project, the result can quickly run out of control. First, the
initial plan of integrating all data sources was not feasible for the limitations of this project.
55

Even after scaling the data entry down to ten years, the result remained too large of a task. Four
years of data resulted as the best option. The spread of four years demonstrates the capabilities
of the software, and gives a viewpoint of long-term planning.

5.2.2 Aerial Images and Processing
While aerial images are valuable to this project and the processing of ortho imagery is a major
step in bringing the images together. Due to the size of the farm, the data collection occurred in
phases as explained in the previous chapter. Because of this separation of data, the processing of
images into ortho mosaic layers processed in batches. While a conclusive product of the farm,
the result was a set of files that do not quite match in their image coloring, and the overlapping
sections are not concrete. What could be an interesting and informational overview of the farm
is choppy in certain places, drawing focus away from the data contained within. To try to
remedy this issue, the project attempted to run a batch process of all the images collected during
the individual flights. To reduce runtime on the processing, I removed the variable of
Drone2Map processing 3D products, and instead focused only on processing the singular ortho
mosaic. The weight of the image processing was not possible for the PC that I was running on.
The computer-processing unit (CPU) was not able handle the stress of processing such a heavy
load. If this project continues, the processing of images would need to run on a computer with a
higher CPU power, or a level off acceptable error will need to be established.
Another technological hurdle of this project was the technical difficulty of connection between
the DJI GPS application and the standard vision system app DJI Go4. During the first attempt of
flight procedures, there were no issues with either application. After establishing connection,
the flight path loaded properly and the waypoints added to the path. Seventy-four images
56

collected without incident. Returning to the farm on a different day to continue to collection,
both apps opened to begin flight preparations, loaded the flight path in to the GSP app, and set
up the UAV to fly. After take-off, standard procedure is to switch from the GO4 App, to the
loaded GSP app, and relinquish control so to the automated flight could take place. After
switchching over to the GSP app, the connected previously seen before takeoff was gone. The
aircraft lost connection to the app. In an attempt to remedy the situation, I switched back to the
GO4 app, only to realize the application disconnected there as well. This resulted in no visual
communication with the UAV. Operational controls with the remote controller remained intact,
but the ability to see the forward vision system, and camera did not exist. At the point, I
scrubbed the mission in the field. Using manual maneuvers, the UAV landed safely and
dismantled for proper technology inspection after I returned from the field. A full inspection
and update verified the UAV was in full working order. It was not clear, and remains unclear, as
to why this error occurred. As a solution to this issue, all flights continued at the same
parameters as the automated flight, but the flight pattern became difficult to stick with. While
navigation in a grid-like pattern is possibly with a UAV, automation is much more precise and
generates a consistent product, instead of a wobbly grid of flight passes. This error also causes
gaps in the imagery. While it may seem like the entire section of farm received coverage during
a flight, it is not fully possible to know what was collected and what was not until the images
processed.

This projects intention was to introduce the principals and concepts of geospatial technology
into the farm management practices at the Homer C. Thomson Research Farm. Because there
was no system of record in place beyond these paper maps, integration of this project has shown
57

deficits in the current system. It its current practice, there is no dedicated steward for this and
any farm specific data within the University. If this project, in its intention, is meant to be a
stepping stone for geospatial technology and precision farming at the Thompson Research
Farm, or any Cornell-run farm, a system of data management and records keeping needs to be
set in place to maintain data quality across the University. Currently, the university data
stewards focus solely on campus planning, transportation, utility and real estate data. Individual
colleges within the university do not have any data stewardship programs. Single farm
agricultural data is stored and maintained by the individual farms, but if a system is set in place
that focuses on brining farm management data to a head, a single managing body should be set
up to maintain the relationship between legacy farm data, and geographical data that can be of
use to both researchers and farm managers.

5.3 Future Projects

Beyond the Homer C. Thompson Research Farm, Cornell manages five research farms,
with an additional three hundred twenty five acres in the immediate vicinity of campus that
compromise another eleven small farms. Each of these farms contains their own unique set of
farm legacy data, and unique geographic features. Figures 16 and 17 represent current maps of
Musgrave Research Farm, and the Willsboro Farms respectively. If it is determined a success,
and ran until its complete integration, this project can be a template for the rest of Cornell-owned
agricultural assets to move to a similar system. Even with limited data, the relationship between
institutional knowledge and geography shows a different perspective to staff and researchers
looking to develop a better understand the area they are working in. It is, however, completely
feasible that this project will only be completed to a certain extent at the Thompson Research
58

Farm, and never be integrated at any other agricultural location. The reality of projects such as
these remains that after the project finishes from a research perspective nothing happens. Waived
are all other expectations and projections of other projects due to interest changes. It is my hope
that the value and knowledge that comes from even a small-scale integration inspires others at
Cornell to look into the data behind the large agricultural network, and develop a better
understanding of geographic and data relationships.

5.4 Conclusions
While small-scale implementation was not my original intention for this project, the
opportunities and results far outweigh any hurdles experienced. With the data uploaded to
ArcGIS Online, farmers at the Homer C. Thompson Research Farm now have the opportunity to
test and discuss the value of geospatial integration with researchers and other staff on the farm.
This data will continue to live on the Cornell ArcGIS Server, and modified as the project
progresses beyond the scope of this paper.

59

Figure 16. Cornell Musgrave Farm
60

Figure 17. Cornell Wilsboro Farm

61

REFERENCES

Auernhammer, Hermann. "Precision Farming — the Environmental Challenge." Computers and
Electronics in Agriculture 30, no. 1-3 (2001): 31-43. doi:10.1016/s0168-1699(00)00153-8.

Bisht, P., P. Kumar, M. Yadav, J.S Rawat, M.P Sharma, and R.S Hooda. "Spatial Dynamics for Relative
Contribution of Cropping Pattern Analysis on Environment by Integrating Remote Sensing and
GIS." International Journal of Plant Production 8, no. 1 (January 2014). doi:1735-6814.

Brock, Debra, Peter Knapp, Mike McMahon, Dan Weddle, and Brian Young. "Towns of Homer, Preble
& Scott Agriculture & Farmland Protection Plan." American Farmland Trust, 2014.

Bruin, Sytze De, Peter Lerink, Inge J. La Riviere, and Bas Vanmeulebrouk. "Systematic Planning and
Cultivation of Agricultural Fields Using a Geo-Spatial Arable Field Optimization Service: Opportunities
and Obstacles." Biosystems Engineering 120 (2014): 15-24. doi:10.1016/j.biosystemseng.2013.07.009.

Casalegno, Stefano, Richard Inger, Caitlin Desilvey, and Kevin J. Gaston. "Spatial Covariance between
Aesthetic Value & Other Ecosystem Services." PLoS ONE 8, no. 6 (2013).
doi:10.1371/journal.pone.0068437.

Environmental Systems Research Institute, Inc. 1995

"Farm Resources for Cornell Researchers." Cornell University Agricultural Experiment Station. Accessed
September 17, 2018. https://cuaes.cals.cornell.edu/farms/maps/.

Feizizadeh, Bakhtiar, and Thomas Blaschke. "Land Suitability Analysis for Tabriz County, Iran: A Multi-
Criteria Evaluation Approach Using GIS." Journal of Environmental Planning and Management 56, no. 1
(2013): 1-23. doi:10.1080/09640568.2011.646964.

Hart, John Fraser. The Rural Landscape. Baltimore, MD: Johns Hopkins University Press, 1998.

Inspire 2 User Manual. PDF. Shenzhen, China: DJI, July 11, 2017.

Jackson, T. "Vegetation Water Content Mapping Using Landsat Data Derived Normalized Difference
Water Index for Corn and Soybeans." Remote Sensing of Environment 92, no. 4 (2004): 475-82.
doi:10.1016/j.rse.2003.10.021.

Mueller, Tom, and Gretchen F. Sassenrath. GIS Applications in Agriculture. Boca Raton: CRC Press,
2015.

"Natural Resources Conservation Service." What Is Soil Conservation? | NRCS. Accessed September 03,
2018. http://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/ref/?cid=nrcs142p2_054242.

"Natural Resources Conservation Service." Web Soil Survey| NRCS. Accessed September 17, 2018.
https://websoilsurvey.sc.egov.usda.gov/App/WebSoilSurvey.aspx.

Neményi, M., P.á. Mesterházi, Zs. Pecze, and Zs. Stépán. "The Role of GIS and GPS in Precision
Farming." Computers and Electronics in Agriculture 40, no. 1-3 (2003): 45-55. doi:10.1016/s0168-
1699(03)00010-3.

62

Pecze, Zs., M. Neményi, P.á. Mesterházi, and Zs. Stépán. "The Function of the Geographic Information
System (GIS) in Precision Farming." IFAC Proceedings Volumes 34, no. 26 (2001): 15-18.
doi:10.1016/s1474-6670(17)33625-x.

Popowitch, Frank M. Developing an Internet-based Geographic Information System for the Cornell
Emergency Response Community. 2010.

"Research Farms." Sustainable Greenhouses and Growth Chambers | Cornell University Agricultural
Experiment Station. Accessed September 15, 2018. https://cuaes.cals.cornell.edu/farms/.

Underhill, Jimmy. "Using DJI GS Pro to Plan Agricultural Survey Missions." Agribotix FAQ. March 5,
2018. Accessed September 28, 2018. https://agribotix.zendesk.com/hc/en-us/articles/115000814654-
Using-DJI-GS-Pro-to-Plan-Agricultural-Survey-Missions.

Weerakoon, Kgpk. "Suitability Analysis for Urban Agriculture Using GIS and Multi-Criteria
Evaluation." International Journal of Agricultural Science and Technology 2, no. 2 (2014): 69. Accessed
October 28, 2016. doi:10.14355/ijast.2014.0302.03

Abstract (if available)

Abstract Over time, the methods and technologies by which we produce and harvest our food have advanced. Large corporations are quick to adopt new technologies and processes, but smaller farms can struggle to see the value in pursuing advanced technologies for farm management. Development of a streamlined protocol for introducing geospatial technology at the individual farm level can help prioritize operations, and help develop long-term operational plans. While the benefits of integrating GIS software and tools are apparent to corporate farm managers, or agricultural economists, they are not always as apparent or easily accessible to small farmers. ❧ Using the research farm in Freeville as a case study, a developmental framework for other farmers at Cornell University and around Tompkins County for small-scale geospatial data integration will develop. With a focus on easy-to-obtain datasets, the procedures outlined on this paper will articulate in a way that non-geospatial data users can understand and build upon. ❧ This paper will examine the possible benefits of implementing GIS technology in small farming communities of like that of the Homer C. Thompson Research Farm and discuss how it can improve the visualization and management of small farms. While the overall impact of introducing geospatial technology at the small farm level is not quantifiable in this paper, understanding what data is available, and its impact on farm operations, can be beneficial in the long-term planning and management of small farms.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Creating a Web GIS to support field operations and enhance data collection for the Animal and Plant Health Inspection Service (APHIS)

PDF

Implementing spatial thinking with Web GIS in the non-profit sector: a case study of ArcGIS Online in the Pacific Symphony

PDF

Utilizing advanced spatial collection and monitoring technologies: surveying topographical datasets with unmanned aerial systems

PDF

Using GIS to predict human movement patterns in complex humanitarian emergencies: a test case of the Syrian Conflict

PDF

Comparative 3D geographic web server development: visualizing point clouds in the web

PDF

Alaska Hike Search: designing a mapping application for Alaskan trails and user contributed hazards

PDF

Building a spatial database of biochar research and practice with Web-GIS

PDF

Building sustainable Installation Geospatial Information and Services (IGI&S) programs: a program management framework of capacity building strategies

PDF

Wetland mapping and restoration decision making using remote sensing and spatial analysis: a case study at the Kawainui Marsh

PDF

A methodology for a real estate blockchain application utilizing geographic information systems (GIS)

PDF

Using the Digital Shoreline Analysis System (DSAS) to analyze changes in shoreline position caused by seawalls along a section of Oregon's coast

PDF

Development of a Web GIS application to aid marathon runners in the race selection and planning process

PDF

Beyond visual line of sight commercial unmanned aircraft operations: site suitability for landing zone locations

PDF

An application of aerial drones in high definition mapping for autonomous vehicles

PDF

A Web GIS application for airport pavement management

PDF

GIS-aided production of equipment locator maps for Metro rail maintenance and support

PDF

Using Landscape Integrity Index to evaluate the cumulative impacts of BLM resource management programs

PDF

A comparison of two earthquake events in the City of Downey: the Puente Hills and Whittier faults at 7.0 and 6.8 magnitudes

PDF

Using GIS to perform a risk assessment for air-transmitted bioterrorism within San Diego County

PDF

Collecting and managing VGI infrastructure assessments in support of stability operations

Asset Metadata

Creator Colomaio, Mary Catherine (author)

Core Title Integrating GIS into farm operations at the Homer C. Thompson Research Farm in Freeville, New York

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 11/07/2018

Defense Date 10/17/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Agriculture,ArcGIS,OAI-PMH Harvest,plots,precision farming

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Fleming, Steven (committee chair), Marx, Andrew (committee member), Wilson, John P. (committee member)

Creator Email colomaio.mary@google.com,colomaio@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-103956

Unique identifier UC11676693

Identifier etd-ColomaioMa-6937.pdf (filename),usctheses-c89-103956 (legacy record id)

Legacy Identifier etd-ColomaioMa-6937.pdf

Dmrecord 103956

Document Type Thesis

Format application/pdf (imt)

Rights Colomaio, Mary Catherine

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA