RFI 2026002 - LULC - FINAL.pdf

Request for Information

Land Use Land Cover Automation

DEP RFI # 2026002

DEP RFI # 2026002 1 of 12 (Revised 04.25.22)

REQUEST FOR INFORMATION

Land Use Land Cover Automation

DEP Posting Number: DEP RFI # 2026002

The Florida Department of Environmental Protection (“Department or DEP”) is issuing this

Request for Information (RFI), as defined in Section 287.012(22), Florida Statutes (F.S.), to

Interested Entities concerning certain commodities and services relating to methods, processes, techniques, technologies, or data that could improve the development of the Land Use Land

Cover dataset, including the use of artificial intelligence, machine learning, object detection, and other technologies. Please note that under Section 287.012(22), F.S., responses to this request are not offers and will not be accepted by the Department to form a binding contract.

Interested Entities can view and download this RFI from the Vendor Information Portal (VIP) website at https://vendor.myfloridamarketplace.com/. Once at this site, the steps listed below should be followed to access the VIP.

• Click “Search Advertisements”

• Under the “Organization” search field, select the “Department of Environmental Protection”

• Scroll down and click on “Search”

• Click on the applicable RFI number https://vendor.myfloridamarketplace.com/

DEP RFI # 2026002 Page 2 of 12

TABLE OF CONTENTS

SCHEDULE OF EVENTS

Section 1.00 – INTRODUCTION

1.01. Purpose and Scope

1.02. Procurement Officer

1.03. Questions

1.04. Addenda

1.05. Submittal of Response

1.06. Public Records

1.07. RFI Process

SECTION 2.00 – TECHNICAL REQUIREMENTS

2.01. Overview

2.02. Background

2.03. Response

2.04. Minimum Acceptable Standards

SECTION 3. 00 – VENDOR CONTACT INFORMATION

SECTION 4.00 – COST INFORMATION

DEP RFI # 2026002 Page 3 of 12

SCHEDULE OF EVENTS

The following schedule will be strictly adhered to in all actions relative to this posting. The Department reserves the right to make adjustments to this schedule and will notify Interested Entities in the request by posting an addendum on the VIP. It is the responsibility of the Interested Entity to check VIP on a regular basis for such updates.

DATES EVENTS METHOD

August 15, 2025 RFI Advertised Vendor Information Portal https://vendor.myfloridamarketplace.com/

August 29, 2025, 5 pm EST

Questions Submitted in Writing

Submit via email:

Attn: Chris Lowery, Procurement Officer E-mail: Christina.Lowery@FloridaDEP.gov

RFI NUMBER MUST BE IN SUBJECT LINE OF EMAIL

On or about, September 12, 2025

Answers to Questions Posted

Vendor Information Portal https://vendor.myfloridamarketplace.com/

September 26, 2025, 5 pm EST

Response Due

Submit via email:

Attn: Chris Lowery, Procurement Officer E-mail: Christina.Lowery@FloridaDEP.gov

RFI NUMBER MUST BE IN SUBJECT LINE OF EMAIL

– Remainder of Page Intentionally Left Blank – mailto:Christina.Lowery@FloridaDEP.gov mailto:Christina.Lowery@FloridaDEP.gov

DEP RFI # 2026002 Page 4 of 12

Section 1.00 – INTRODUCTION

1.01. Purpose and Scope. The Florida Department of Environment Protection (“Department”) is requesting information from Interested Entities with proven, verifiable, and documented expertise in methods, processes, techniques, technologies, or data for the development of the Department’s Statewide Land Use and Land Cover (LULC) dataset such that the dataset is developed timely, consistently and accurately within the State of Florida. This Request for Information (RFI) is being issued strictly for the purpose of gaining knowledge of the products and services available on the market.

THIS IS NOT A REQUEST OR INVITATION FOR COMPETETIVE BIDS, PROPOSALS OR REPLIES, AND NO

CONTRACT WILL BE AWARDED IN RESPONSE TO SUBMISSIONS.

If, as a result of the RFI, the Department decides to pursue contract(s) to provide any of the services mentioned and Interested Entities capable of providing such services are identified, a competitive procurement package may be developed and issued at a later date.

1.02. Procurement Officer.

Chris Lowery, Procurement Officer Bureau of Procurement & Contracting Services Carr Building, Room 215 Florida Department of Environmental Protection 3800 Commonwealth Boulevard, MS#93 Tallahassee, Florida 32399-3000 Telephone Number: (850) 245-2380 Christina.Lowery@FloridaDEP.gov

1.03. Questions. Any questions from vendors concerning this RFI shall be submitted in writing, identifying the submitter and RFI number, to the Procurement Officer no later than the time and date specified in the Schedule of Events. All inquiries must be submitted by E-mail. All questions and answers will be posted on the VIP. It is the Vendor’s responsibility to periodically check the VIP. The Department bears no responsibility for any delays, or resulting impacts, associated with a Vendor’s failure to obtain the information made available through the VIP.

1.04. Addenda. If the Department finds it necessary to supplement, modify, or interpret any portion of the RFI documents, a written “Addendum” will be posted on the VIP. It is the responsibility of the Vendor to be aware of any Addenda that might have a bearing on their response.

1.05. Submittal of Response. Responses are due in accordance with VIP and Schedule of Events.

Vendors’ responses must be in writing and submitted electronically via email. Responses must be emailed to the Procurement Officer (Section 1.02). The naming convention for the subject line within your email is:

Subject Line: Your Company Name – DEP RFI # 2026002

Once received, all RFI responses and attachments shall become the exclusive property of the Department and will not be returned. The Department is not liable for any of the costs incurred by the Vendor in preparing and submitting a response.

The response package must contain a title page including the below contact information:

• Vendor/Company Name;

• Contact Name and Title;

DEP RFI # 2026002 Page 5 of 12

• Address; and

• Phone Number and email address.

To ensure a clear understanding of the products and service, the Department may use this RFI as the basis for requesting demonstrations. These demonstrations will be at the department’s discretion and at no cost to the Department.

1.06. Public Records. Any material submitted in response to this RFI will become a public record pursuant to Chapter 119, Florida Statutes (F.S.), once it is shared with the Department. Any claim of confidentiality is waived upon submission, unless addressed as set forth below.

Disclosure and Ownership of Response Contents: A Vendor’s response to this RFI shall be a public record and subject to production, disclosure, inspection and copying consistent with the requirements of Chapter 119, F.S. All information in a Vendor’s response (including, without limitation, technical and price information) will be a matter of public record, subject to the provisions of Florida’s Public Records Act, Chapter 119, F.S., regardless of copyright status. Submission of a response to this RFI shall constitute a waiver of any copyright protection which might otherwise apply to the Department’s production, disclosure, inspection and copying of such response and contract, or any part thereof, except those parts asserted to be exempt under Chapter 119, F.S. The response, upon submission shall be the property of the Department (except those parts asserted to be exempt in the manner set forth below), and the Department, in its sole discretion, shall have the right to use, reproduce, and disseminate the response. The Department reserves the right to use any and all information contained in a response received to this RFI.

Any content submitted to the Department which is asserted to be exempt under Chapter 119, F.S., shall be set forth on a page or pages separate from the rest of the response, and clearly marked “exempt,” “confidential,” or “trade secret” (as applicable), with the statutory basis for such claim of exemption, confidentiality, or trade secret specifically identified in writing on each and every such page. Failure to segregate and so identify any such content shall constitute a waiver of any claimed exemption, confidentiality, or trade secret as applied to the portion of the response or other document in which the content is set forth.

Per Section 287.012(22), F.S., “Responses to these requests are not offers and may not be accepted by the agency to form a binding contract.” Vendors submitting information to this RFI are not prohibited from responding to any related subsequent solicitation. The Department reserves the right to use or reject any information supplied in response to this RFI.

1.07. RFI Process.

The Department will review and analyze information received in response to this RFI to determine the feasibility of issuing a competitive solicitation for these products/services. Any request for cost information received will be used solely to gain a perspective of the potential budgetary magnitude.

Responses to this RFI will be reviewed for informational purposes only and will not result in the award of a contract. Submitting a response to this RFI shall not disqualify a vendor for responding to any related subsequent solicitation.

DEP RFI # 2026002 Page 6 of 12

SECTION 2.00 – TECHNICAL REQUIREMENTS

2.01. Overview. DEP creates and uses the Statewide Land Use and Land Cover (LULC) dataset as a critical component for the proper modeling of watersheds to determine nutrient and pollutant loading for developing total maximum daily loads and basin management action plans and associated allocations, among other uses. Water Management Districts (WMDs) use LULC for their modeling and planning efforts as well. LULC classifications are assigned by the South Florida WMD, Southwest Florida WMD, and St. Johns River WMD as well as DEP for the Northwest Florida WMD and Suwannee River WMD. DEP combines the individually developed LULCs for the state’s five (5) water management districts into a single statewide dataset, which is updated periodically when new data is available, but is generally refreshed on three (3) year intervals statewide. Each agency uses the 1999 Florida Land Use, Cover and Forms Classification Handbook and most utilize photo interpretation by trained staff as the classification methodology. This dataset is published online and is used by other agencies, local governments, and private stakeholders.

The Department has embarked on an effort to meet the below specific project objectives:

(1) Make the development of the LULC dataset more consistent statewide;

(2) Establish and implement new data quality checks to improve data accuracy statewide; and

(3) Increase efficiency in data development while reducing administrative burdens, including, where possible, costs.

To that end, the Department seeks to understand and identify potential methods, processes, techniques, technologies, or data that will help achieve those objectives with this crucial dataset.

2.02. Background. The Department contracted with The Balmoral Group to complete an evaluation of its use of LULC dataset, which resulted in a Final Report, see Attachment 1. The Department additionally evaluated individual LULC codes to determine which Land Use codes were critical to Department or WMD work as well as where Land Cover codes would be beneficial, see Attachment 2.

The differences between land use and land cover are crucial for the Department’s work and there should be an important distinction for entities responding to this RFI or any future solicitation. The

Department’s work typically relies heavily on the use of the land, not on land cover. For example, the

Department needs to know if a structure is a wastewater facility and not simply identify it as a structure.

Notwithstanding, the Department has identified several land uses where a separately classified land cover would provide value.

In order to meet the project objectives identified above, the Department has identified two areas of improvement that may be possible with new, innovative and cutting-edge technologies: imagery or remote sensing data and data automation.

A. Imagery or Remote Sensing Data

The Department recognizes that significant advancements have been made in imagery acquisition, including through drones and satellites, as well as traditional methods, such as airplanes. Other remote https://geodata.dep.state.fl.us/datasets/2f0e5f9a180a412fbd77dc5628f28de3_3/explore https://floridadep.gov/owper/water-policy/content/water-management-districts https://fdotwww.blob.core.windows.net/sitefinity/docs/default-source/content/geospatial/documentsandpubs/fluccmanual1999.pdf?sfvrsn=9881b4d0_0 https://fdotwww.blob.core.windows.net/sitefinity/docs/default-source/content/geospatial/documentsandpubs/fluccmanual1999.pdf?sfvrsn=9881b4d0_0

DEP RFI # 2026002 Page 7 of 12 sensing platforms may additionally provide means to classify land uses and land covers on a statewide level to assist the Department in achieving its objectives.

Currently, the Department solely uses aerial imagery with a resolution of six (6) inches for most areas of the state; the Department generally would like to keep the same resolution or better for any aerial imagery. In no case should the resolution options exceed one (1) foot; as the Department is evaluating automation options, however, higher resolution is preferred. It is not expected that the Department will need imagery with higher resolution than three (3) inches.

The purpose of this RFI is to understand:

The type, quality, and refresh frequency associated with imagery or other relevant remote sensing data acquisition;

Data collection methods;

Data quality assurance and control standards; and

Where possible and available, cost estimates or ranges and cost structures (such as subscription-based cost structures).

B. Data Automation

The Department recognizes that significant advancements have been made in identifying land surface features using automation, including, but not limited to, Artificial Intelligence (AI) and Machine Learning.

Notwithstanding, the Department is aware that these tools may have limitations that should be carefully considered. The purpose of the LULC evaluation was to better understand the minimum standards needed by the Department in its use of the LULC dataset so that the appropriate tools may be evaluated and, where appropriate, implemented.

The purpose of this RFI is to understand:

1. Existing or cutting-edge technologies capable of completing this work and whether this work could be completed through one technology/service or multiple (for example, whether the Department should pursue multiple solicitations to achieve each of the Minimum Acceptable Standards identified below).

2. Control Standards that should be included in a future solicitation. Specifically, the Department is seeking to understand what performance measures can be identified in a future solicitation that will give the Department a high confidence in the data output. This would be particularly valuable to any AI-based or other automation tools. (For example, should the Department utilize performance-based metrics in the solicitation and, if so, what might those include.)

3. The estimated time necessary to complete the implementation of the described method, process, technique, technology, or data on a statewide basis.

4. Where possible and available, cost estimates or ranges and cost structures.

2.03. Response. The Department is interested in information that will result in a greater understanding of the services, data, and estimated costs associated with updating the statewide LULC dataset.

Responses should be based upon existing or cutting-edge technologies to support the specific project objectives.

DEP RFI # 2026002 Page 8 of 12

All responses to this RFI must contain the following information:

• A brief description of the Interested Entity’s experience and credentials that show the Entity is qualified to provide the Department with the information requested by this RFI.

• A brief description of the methods, processes, techniques, technologies, or data proposed by the Interested Entity to assist the Department with its LULC objectives and meet at least one of the minimum acceptable standards outlined in section 2.04 and that addresses the purposes described in 2.02, above.

• For vendors proposing the use of AI, provide information about how the Department may solicit while ensuring sufficient data security as well as which learning environment the AI may leverage.

• A statement as to whether the methods, processes, techniques, technologies, or data described by the Interested Entity is in use by other governmental or non-governmental entities with details of use case where possible.

• Any additional information that the Department should be aware of or should consider. The

Department welcomes vendors and experienced entities to expand on any relevant topics even where not specifically described in this RFI.

2.04. Minimum Acceptable Standards. The Department seeks to identify what method, process, technique, technology, or data may be available to aid the Department in achieving its specific project objectives through future solicitation if in the best interest of the state.

The Department has identified minimum standards to be considered for future solicitation.

1. Imagery or Remote Sensing Data Acquisition: The Department would be interested in information about type, quality, and refresh frequency associated with imagery or other relevant remote sensing data acquisition that would support the Department’s LULC objectives. The current data is updated approximately every three (3) years, with some exceptions, and more frequent data may be valuable to broader audiences, those the Department would evaluate cost effectiveness of a more frequent dataset. At a minimum, the Department asks that the data:

a. Be collected statewide in a single season (current imagery is collected by county over a period of three (3) years through the Department of Revenue for most);

b. Be collected in Winter (unless seasonal tree cover is not applicable to the remote sensing technology);

c. Have consistent color and/or quality statewide;

d. Have resolution that would support automation tools identified below to a minimum mapping unit of 0.5 acres; and

e. Must be able to be registered (geo-referenced) to a base map provided by the state agency, such as Department of Transportation (DOT).

2. GIS Spatial Quality Check Automation. The Department would be interested in methods that help compare authoritative data sets to the LULC data set for automated or semi-automated quality control and correction. The Department would like:

a. Automated tools to efficiently perform quality checks over large datasets that, once verified, could be corrected/updated in the dataset by qualified reviewers. For example, the Department houses point data for wastewater facilities. Wastewater facilities are a

DEP RFI # 2026002 Page 9 of 12 land use polygon in the LULC dataset. Tools that can “flag” a polygon based on another dataset’s geolocational data, such as wastewater facility points, would provide great benefit if it could be done through dozens of datasets on a statewide basis efficiently and accurately.

b. Where feasible, the Department would be interested in the automation of corrections to the LULC dataset if such corrections could be verified by qualified reviewers until such time the corrections are trusted as consistently accurate.

Vendors may identify ancillary data sets that it may deem valuable in confirming or correcting LULC classifications.

3. LULC Classification Automation. The Department would be interested in methods that would automate or semi-automate updates to the LULC dataset but is equally interested in understanding potential error rates and the controls that may be put in place to avoid or minimize such errors. Specifically, the Department would be interested in:

a. Land use change detection. Because the LULC has been developed, and with additional edits to the classifications pursuant to #2, above, from one cycle to the next, land use does not change drastically at the statewide level. However, changes can be significant at the local level. For this reason, change detection may be valuable.

b. Feature identification. Consistent with the delineation automation below, automation of feature identification, including boundary comparisons, classification quality checks, and more.

c. Land use delineation automation. Land uses are delineated by different staff at the WMDs and Department. Where possible, these delineations may be corrected to create more consistency statewide. (For example, in delineating a road, how much of the right of way should be included in the polygon, if any, and should/when should medians be within their own polygon? As another example, using elevation changes (as an existing authoritative dataset) between natural features on a slope (transition points) to delineate different water boundaries.

d. Level three (3) or four (4) re-classifications. WMDs and the Department assign different classification levels based on needs. The Department has identified some key level three

(3) or four (4) classifications that would be valuable to its work if consistently implemented statewide. Where possible, either using other datasets (see #2 above) or other means, where a level two (2) could be re-classified at a Level three (3) or four (4) classification in accordance with those unmet needs, automation would be valuable.

e. Land Cover classifications. In nearly all circumstances, the WMDs and Department do not need a land cover classification that is different than the land use classification. However, for some land uses, delineating and classifying distinct land covers would be valuable, particularly if this work was automated. (See Department Spreadsheet) This includes:

i. Potentially delineating some classification codes for pervious and non-pervious surfaces.

ii. Potentially delineating only certain classification codes for a separate land cover code. Specifically, this would be valuable where valuable land uses (such as military) include natural features (such as wetlands).

f. Secondary Land Use or Intersect classifications. Certain land uses are overlapping, such as electrical transmission lines over wetlands or bridges over rivers. Balmoral found that these conflicting land uses were not handled consistently statewide and sometimes resulted in “interrupted” classifications. The Department is considering addressing this by creating secondary land uses that is, a second land use code for the same polygon.

DEP RFI # 2026002 Page 10 of 12

Assigning these secondary land uses automatically would be valuable. Examples include bridges, transmission lines, spray fields, marinas (water side) and similar overlapping land uses. What other structures to the data would you recommend?

DEP RFI # 2026002 Page 11 of 12

SECTION 3. 00 – VENDOR CONTACT INFORMATION

Providers are asked to provide the following information:

CONTACTS

Primary contact regarding this submission? Name:

Email Address:

Phone:

Title or Role in Organization?

Primary contact to schedule Presentation:

(If other than contact above)

Name:

Email Address:

Phone:

Title or Role in Organization?

Company Website URL:

Type of Organization (Corporation, Partnership, etc.):

How long has your company been in this type of business?

Years:

Months:

Location of Project Manager that would serve the Department:

City:

State:

Location of Regional Sales Manager that would serve the Department:

City:

State:

DEP RFI # 2026002 Page 12 of 12

SECTION 4.00 – COST INFORMATION

Providers are asked to provide the following cost information for the information otherwise provided in their response, which may include all or some of the identified project objectives and minimum standards: [Section can be modified, including duplicating the table for more than one project objective identified in the response]

DESCRIPTION

TIMELINE

(If Applicable)

PER/UNIT OF

MEASURE

QUANTITY

ESTIMATED

DOLLAR

AMOUNT

Estimated Support Rates:

Estimated Annual Licensing or Subscription Cost:

Estimated Upgrades and other Maintenance Costs per Year:

Estimated Third Party Licenses or Subscription Costs:

Estimated Cost of Implementation Including Migration Service Fees:

Length of typical implementation:

STATE OF FLORIDA

DEPARTMENT OF ENVIRONMENTAL

PROTECTION

Attachment 1

Land Use Land Cover Data in Florida:

Dataset Analysis and Findings on Data Use and Data Refinements

Final Report

RFI 2026002

The Balmoral Group

Land Use Land Cover Data in Florida:

Dataset Analysis and Findings on Data Use and Data Refinements

Final Report

The Balmoral Group, LLC 165 Lincoln Avenue

Winter Park, FL 32789 Phone: 407-629-2185

Dataset Analysis and Summary of Regional Differences

Prepared by Prepared for

The Balmoral Group Florida Department of Environmental Protection

Web - www.balmoralgroup.us Web - https://floridadep.gov/

Head Office

165 Lincoln Avenue

Winter Park

Florida, 32789, USA

Phone 1 407 629 2185

3900 Commonwealth Boulevard

Tallahassee, Florida 32399-3000

Tallahassee Office

113 S Monroe Street

Tallahassee

Florida, 32301, USA

Phone - +1 850 201 7165

Seattle Office

1201 2nd Ave

Seattle

Washington, 98101, USA

Phone - +1 206 600 4197

Sydney Office

Suite 1, Level 2

210 George St

Sydney, 2000, Australia

Phone +61 2 9247 9670 http://www.balmoralgroup.us/ https://floridadep.gov/

Acronyms and Abbreviations BMAP – Basin Management Action Plan

MMU – Minimum Mapping Unit

DEP – Florida’s Department of Environmental Protection

Districts – Florida’s water management districts (WMD)

FDACS – Florida Department of Agriculture and Consumer Services

FDOR – Florida Department of Revenue

FDOT – Florida Department of Transportation and

FHD – Florida Hydrography Dataset

FNAI – Florida Natural Areas Inventory

FSAID – Florida Statewide Agricultural Irrigation Demand – a project and datasets of FDACS

LULC – Land Use Land Cover

NHD – National Hydrography Dataset

SME – subject matter expert

NWI – National Wetlands Inventory

NAIP – National Ag Imagery Program

USACE – United States Army Corps of Engineers

NWFWMD – Northwest Florida Water Management District

SFWMD – South Florida Water Management District

SJRWMD – St Johns River Water Management District

SRWMD – Suwannee River Water Management District

SWFWMD – Southwest Florida Water Management District

Executive Summary Florida Department of Environmental Protection (DEP) retained The Balmoral Group to clearly document how Land Use Land Cover (LULC) data is being developed and refined and to identify all the ways selected agencies in Florida are using LULC data. The purpose of this report is to provide relevant information on

LULC data use and data updates in order to inform possible future changes in how LULC are updated or maintained.

This report provides:

• Summarized findings about LULC data refinement and LULC data use in Florida, based on surveys and interviews and focus group meetings with subject matter experts (SMEs)

• Analysis of the LULC data to provide quantified evaluations of the following 6 topics:

o Feature sizes – how the typical polygon sizes differ by agency and by land use description o Classification levels – how the land use descriptions at Levels 2, 3, or 4 vary across the state o Land Use (LU) and Land Cover (LC) similarity – quantifies how often LU and LC are the same o LU data overlaps – provides descriptions and typical examples of land uses that are by definition overlapping (bridges overlapping roads or waterways and transmission lines overlapping wetlands or agricultural lands, for example) o Critical assets identification – documents the statutorily-defined critical asset types that align with existing LULC descriptions o Agricultural land use – summarizes how the statewide agricultural LULC data differs from

FDACS’ coverage of agricultural lands.

An online LULC data viewer or “dashboard” was developed to allow for additional visualization to support the data summaries provided in this report. The dashboard can be accessed here:

https://datavisual.balmoralgroup.us/FL-LULC-dataviewer.

Recommendations on LULC data update changes While the design of any potential changes in LULC data update processes are beyond the scope of this project, several findings of particular relevance are mentioned for consideration:

- Critical assets spatial data layers from DEP could be used statewide to improve LULC data consistency (updating LU descriptions based on asset type and location of asset). This serves both to provide better representation of critical assets in the LULC data and allows for areas of critical assets to enable more meaningful flood risk evaluations at locations for which critical assets are represented by points.

https://datavisual.balmoralgroup.us/FL-LULC-dataviewer

- Level 3 and 4 LU classifications are often used by LULC data users, but there are numerous distinct LU classifications at those levels – not all of which are required. Providing LULC data users with candidates of simplified LU description schema or collecting data users’ existing crosswalk tables would benefit the design and would allow opportunities for particular LU descriptions that need to be retained to be identified. More generally: LULC data users should continue to be engaged during the LULC data changes design phases.

- Agricultural lands (level 2000) could potentially be simplified in the LU descriptions used for

LULC data updates. Most data users that require detailed crop information relevant to nutrient load or water demand modeling are using FDACS’ FSAID; therefore, this could be an opportunity to classify at level 2 or 1 for agricultural lands in the LULC data to improve updates efficiency. Alternatively, the FDACS FSAID data could be integrated into the LULC data directly.

Both options are discussed in more detail in the Agricultural Land Use section.

Information gathering (surveys, interviews, focus groups meetings) was completed with SMEs who are data users and different SMEs involved in LULC data updates. Florida’s LULC data is a foundational dataset that is used widely for a multitude of purposes. The LULC data stakeholders included in the information gathering here were from DEP, the water management districts (Districts), and FDACS. The following sections explain the key findings related to data refinements and data use of LULC data.

Data Refinement Key Findings:

• Agencies tend to adjust their classification levels (level 2 to 4) based on their data users’ requirements, though most expressed that wetlands and agricultural lands are commonly looked at with the most scrutiny. These land uses take more time in determination of proper codes than other land uses, due to the often-shifting nature of the land. However, there seemed to be opportunities for more direct interaction between LULC data users and data developers to ensure LULC data needs are being met efficiently. Meaning, there did not seem to be established or regular meetings with LULC data users and data developers specifically for the purpose of refining data update/development processes.

• The FDOR/FDOT aerial imagery is the key data source for photointerpretation for LULC data development in Florida for DEP (covering NWFWMD and SRWMD), SJRWMD, and

SFWMD. SWFWMD uses another imagery source, contracting directly for imagery to be flown/collected in their District. Key distinctions among the FDOR/FDOT imagery and

SWFWMD’s imagery collection:

o Both FDOR/FDOT and SWFWMD imagery is flown about every 3 years o However, the FDOR/FDOT has different flight schedules depending on the County;

meaning that within a District the imagery collection dates might be months or years apart for different Counties, making photointerpretation a challenge in some situations.

o SWFWMD collects their own imagery to resolve that challenge of different imagery dates across the District; the imagery data from SWFWMD has very similar flight dates.

o SWFWMD pays for their imagery, while the FDOT/FDOT imagery is already paid by those agencies and is used freely for LULC data updates by the Districts.

• Most agencies noted the inconsistencies in acquisition timeframes for aerial imagery, which can affect the accuracy of LULC coding. The temporal differences make changes in wetlands, water features, and agricultural lands difficult to accurately document. Also, the variability in the color qualities of imagery (resulting from different contractors for different counties) can make photointerpretation more challenging, especially at the intersection of imagery from different sources. Some agencies utilize false color imagery to improve the consistency of imagery interpretation. Most agreed that more consistent and frequent imagery data would be desirable, albeit not at the cost of accuracy. It was noted that there usually are not major LULC changes in a single year, so every 3 years is likely suitable for LULC applications, but more consistent imagery dates across a District would help with photointerpretation.

• Interviews revealed some commonalities in the supporting datasets used to assist with photo interpretation: these included the Department of Revenue (DOR) parcel data

(owner information and use code can be helpful in determining LULC code), the Florida or National Hydrography Dataset (FHD or NHD: to check water or wetland extent), National Wetlands Inventory (NWI), and other imagery sources (Google – including

StreetView, National Ag Imagery Program: NAIP).

• There was discussion with SMEs about the potential use of standardized supporting layers that could be used statewide and could possibly streamline LULC updates. For example, the following data layers might be used statewide to reduce editing time – workflows would need adjustment to incorporate these data layers and maintain a continuous coverage LULC dataset without area overlaps:

o FHD: to represent water and wetlands (polygon datasets) o Roadways – from FDOT and USGS to represent state and local road surfaces (line datasets) o Agricultural lands from FDACS’ FSAID dataset (polygons) o Transmission lines for high voltage electricity lines (included as a linear dataset in

DEP’s critical assets database; maintained/developed by the Department of

Homeland Security Homeland Infrastructure Foundation Level Database – DHS

HIFLD)

There was interest in utilizing some standard statewide datasets to reduce LULC revision areas, but there was not consensus on datasets or process.

• Agencies noted some challenges in LULC classification at the boundaries between districts. This sometimes requires LULC updates to expand across District boundaries to avoid the District-to-District differences in LULC classification that might complicate some use cases having boundaries across Districts. SFWMD and SJRWMD both noted some

LULC updates outside their boundaries to align their LULC with surface watershed basin boundaries as follows.

o SFWMD

▪ Distance outside District: 4 miles to 15 miles beyond the SFWMD boundary; distance varies along boundary based on watershed boundaries

▪ Length of the District boundary for which LULC extend outside the boundary: entire length of District boundary

▪ LULC updates outside boundary based on requests or standard procedure:

standard procedure o SJRWMD

▪ Distance outside District: a few miles to 10 miles beyond the SJRWMD boundary to incorporate any watersheds; distance varies along boundary based on watershed boundaries

▪ Length of the District boundary for which LULC extend outside the boundary: entire length for which watershed boundaries extend beyond

District boundary

▪ LULC updates outside boundary based on requests or standard procedure:

standard procedure

• Suggestions for potential new LU classifications included:

o Solar: Utility-scale photovoltaic (PV) installations o Abandoned structures

• Suggestions for potential new LC classifications included:

o Hurricane damaged natural areas and developed areas o Swimming pools o Residential pervious and impervious surfaces (turf, streets, sidewalks) as separate features rather than being included in residential categories

• Estimated annual costs to update LULC data ranged from about $120,000/yr to

$330,000/yr per agency; most of these costs were internal staff time, with the exception of SFWMD and SWFWMD in which most of the costs were for imagery and contractual services.

The following table summarizes annual costs for LULC data updates:

Table 1. Annual update costs summary by agency

Agency Total Annual Cost**

Data Costs Contractor Costs Internal staff costs

DEP* $150,000 ŧ

$150,000

SFWMD $119,000 ŧ $92,000 $27,000

SJRWMD $125,000 ŧ

$125,000

SWFWMD $337,000 $242,000 $88,000 $7,000

* Note: DEP LULC updates cover both SRWMD and NWFWMD.

**Note: Annual costs based on agency costs provided as annual amounts or adjusted to annual amounts based on LULC update duration (example: 3-year cycle cost total divided by 3 for annual costs).

Imagery costs associated with DEP, SFWMD, and SJRWMD are incurred by the FDOR for flights and by FDOT for data storage.

• Relevant to potential data update process changes: Manual photo interpretation is used exclusively by all SMEs except one. For that exception, machine learning has been explored, but manual photo interpretation was still used for about 80% of their LULC development. All agencies are exploring or planning to explore how to include some type of automation of LULC data refinement, but none have developed any detailed information based on vendor input or other information sources. However, some data update personnel expressed concerns about how the potential inaccuracies may reduce some benefits of “time saved” via automation. Staffing costs have been noted as an important reason why agencies are exploring data updates automation.

Major advantages noted of potential process changes: improved consistency in LU classifications across districts, and reduced time between LULC update cycles. It is expected that automation of some parts of LULC data updates could reduce staff time commitments, but there was not a clear indication if there would be a cost increase or decrease overall. It was noted that most automated imagery classification systems have only 10 or 15 categories (based on District staff exploration in recent years of ESRI’s and other GIS tool LULC classification tools; not based on vendor solicitation directly). There were no data developers directly using automated classification tools. It was recommended by some data update experts that automated processes could potentially provide the Level 1 land use or provide change detection areas only, and manual interpretation could proceed to categorize to the typical LULC level of detail.

Data Use Key Findings:

• Differences in LULC classifications across district boundaries require dataset adjustments for some applications. For DEP, this is particularly relevant for nutrient load model applications. For the Districts, these cross-district LULC data differences result in some districts updating LULC beyond their boundaries to enable the LULC to be more readily usable by district personnel working on data uses near or across boundaries. See findings above: SFWMD and SJRWMD both routinely extend their LULC data updates beyond

District boundaries where needed to align with surface water basin boundaries.

• DEP is using the LULC data statewide and across district boundaries, while the water management districts are typically using LULC just within their district – though there are some exceptions for water supply plan regions that span multiple districts. There are numerous other LULC data stakeholders not included in this scope who are regular users of LULC data.

• Usage frequency of the LULC data for the SME participants in this study was generally about 2 to 3 times per month, with some personnel using LULC daily and others using it just once or twice per year.

• Data use/applications were common for both landscape-scale (basin or planning region) and for site-specific uses (property level assessment). Common applications of LULC data are (Table 2):

Table 2. LULC data use cases summary

Data Use Reported Agency

Water supply modeling: hydrologic models that estimate evapotranspiration and groundwater recharge and other hydrologic processes

All five districts

Water quality modeling: nutrient load estimations DEP, SJRWMD, SRWMD

Flood risk modeling: watershed management plan models SWFWMD, SJRWMD

Watershed prioritization for updating watershed models (age of model and % of LU change)

SWFWMD

Data Use Reported Agency

Permit review processes for WUPs or ERPs All five districts

Project locations and environmental lands assessments:

supports the identification of project locations (for water supply or water quality projects) and the evaluation of environmentally sensitive lands (potential acquisitions or easements)

All five districts

Critical asset area coverage for resilience planning and modeling

(for review and exploratory purposes: LULC areas not used statewide for critical assets spatial representation)

DEP

Note: these are LULC data uses by DEP, the water management districts, or FDACS, as those were the subject matter experts convened in this project. There are numerous other LULC data uses in Florida not accounted for in this assessment.

• The majority of data users indicate that level 3 LU classifications are needed for at least some land use assessments. Level 3 and 4 classifications were preferred by data users most commonly in agricultural lands and wetlands and water features classifications.

Water and wetlands features were consistently mentioned as features in which the most detailed descriptions and spatial representation are needed, as these features are relevant for permitting and other applications. It should be noted that while level 3 and

4 LU descriptions were described as sometimes needed by data users, it is evident from review of sample crosswalk tables that in many cases relatively few of the available level 3 descriptions are used. Meaning, while there are over 130 unique level 3 land use descriptions, only 4 or 5 of those might be used at level 3, while the rest might be used based on level 1 or 2 land use descriptions.

• Between surveys and interviews, the majority of users saw the current frequency of updates (3 years on average) as sufficiently frequent, but general consensus was that if more frequent updates were possible, that would be beneficial. There was a consistent preference for accuracy over frequency; most users would rather not increase update frequency if that resulted in more data uncertainty or classification errors.

• Most users supplemented LULC data with multiple other imagery data sources for verification and supporting information. Some examples included DOR parcel data, soil data, historical LULC data, the NHD/FHD, elevation data, and Google Maps or StreetView for additional imagery review assessment in some cases.

• Several users indicated that the addition of impervious surfaces – possibly as a separate data layer – would benefit their modeling efforts.

• Relevant to potential data update process changes: LULC data users indicated a preference to maintain the current classification levels and names, as these are linked to numerical values for various models (runoff, groundwater recharge, nutrient loading, and other models). However, not all unique LU descriptions are needed for most data uses;

it would likely benefit data users to provide them with candidate classification schemes

(if these are developed) during any potential design stage of LULC data updates.

Additionally, some users expressed concern about potential misclassifications from an automated or semi-automated data update process. Data users welcomed any automation that might improve data update frequency, but had some reservations about potentially increased classification errors or changes in the LULC classification system. As noted from the data updates SMEs, a major advantage of potential LULC update process changes: improved consistency in LU classifications across districts.

TABLE OF CONTENTS

Acronyms and Abbreviations

Executive Summary

Recommendations on LULC data update changes

Data Refinement Key Findings:

Data Use Key Findings:

Feature Size Evaluation

Classification Levels Assessment

Land Use and Land Cover Differences

Land Use Data Overlaps

Critical Assets Identification

Agricultural Land Use

FL LULC Data Viewer

Appendix A – Interview, Focus Group, and Survey Summaries

LULC Data Refinement

LULC Data Use

Appendix B – Supporting Information

LULC Example Crosswalk Table

Critical Assets Data Sources and Geometry Type

Feature Size Evaluation In surveys and interviews with data users, the typical feature sizes or specified minimum mapping units

(MMUs) were not noted to be hampering LULC data applications. Most data users indicated that the spatial resolution or size of features were suitable for their applications, with smaller features often needed or preferred for water or wetland features.

The following summary and figures quantify feature sizes around the state. While LU codes in the 2000 to 7000 range are more likely to vary around the state as croplands and natural areas have more variability, specific land uses in the 1000 (Urban and Built-up) and 8000 (Transportation, Communication, and Utilities) ranges could be expected to have more similarities across the state in terms of average feature sizes. The online LULC data dashboard, https://datavisual.balmoralgroup.us/FL-LULC-dataviewer, was prepared to facilitate feature size comparisons for particular land use descriptions or collectively for level 1 or level 2 land use codes.

Shown below is a comparison of average feature size acreage by agency for the 1000 and 8000 level 1 codes (Figure 1). There are substantial differences between some districts in average feature size for both 1000 and 8000 levels, with average feature areas around 9 to 21 acres in SRWMD, NWFWMD, and

SJRWMD for 8000 level and 62 to 75 acres in SFWMD and SWFWMD for 8000 level (see Figure 1). For level 2 LU codes for comparison within 8000 level codes, there is more similarity in feature sizes by district for the 8200 LU code (Communications): average feature size ranges from 1 to 10 acres. For transportation (8200): average feature sizes are similar in SRWMD and NWFWMD (about 40 acres), but in SJRWMD, SWFWMD, and SFWMD, feature sizes are much larger (about 200 acres on average). This is likely representative of expected roadway size differences in those districts (Figure 2).

Figure 3 illustrates typical size differences across land uses and districts for 1100 = Residential, Low

Density, 1200 = Residential, Medium Density, and 1300 = Residential, High Density land uses. Similar to the findings for 8000 level codes, average feature sizes for 1100, 1200, and 1300 are similar in SRWMD and NWFWMD (about 20 acres), while in SJRWMD, SWFWMD, and SFWMD, residential feature sizes are moderately larger: around 50 acres depending on density group.

It is expected that for level 3 LU codes that would typically be categorizing particular properties, there would be more similarity in feature sizes across districts. This is true in many instances; for example, LU

= 8340 (Sewage Treatment or Wastewater Treatment) shows similar average feature sizes (about 20 acres) across the 4 districts using 8340 (see Figure 4. Note: SWFWMD does not categorize most LU at level 3). Similarly, for 8120 (Railroads), average feature sizes are nearly the same across districts, with

SRWMD showing somewhat larger average feature size (Figure 5).

In surveys and interviews with data users, the typical feature sizes or specified MMUs were not noted to be hampering LULC data applications. Most data users indicated that the spatial resolution or size of features were suitable for their applications.

Figure 1. Average feature sizes for 1000 (Urban and Built up) and 8000 (Transportation, Communications, and

Utilities)

Figure 2. Average feature sizes by district for 8100 = Transportation, 8200 = Communications, and 8300 =

Utilities LU codes.

Figure 3. Average feature sizes by district for 1100 = Residential, Low Density, 1200 = Residential, Medium

Density, and 1300 = Residential, High Density LU codes

Figure 4. Average feature size and total acres for each district for LU = 8340; note SWFWMD does not typically classify LU at level 3

Figure 5. Average feature size and total acres for each district for LU = 8120; note SWFWMD does not typically classify LU at level 3

Classification Levels Assessment There are 218 unique LU codes in the LULC dataset: 42 at level 2, 137 at level 3, and 39 at level 4.

Based on data user surveys and interviews, level 3 LU descriptions are commonly needed for LULC data users, but not for applications by SWFWMD data users. LULC data users were surveyed to identify what level of LU descriptions are usually needed for their uses of LULC data. Users were asked if they

Sometimes, Always, or Never need Level 2 or Level 3 or Level 4 LU descriptions, and they were provided a link to spreadsheet listing all Level 2, 3, and 4 LU descriptions. A response of “Sometimes” indicates that a user does not need all unique LU descriptions at a particular level, rather they need some of them.

For example, for some SJRWMD LULC data uses, the level 2000 (Agriculture) codes needed by SJRWMD require 14 Level 3 land use descriptions, while all the other agriculture codes could be coded as “General

Agriculture” for their purposes. See Appendix Table 7 (LULC Example Crosswalk) for the agricultural land cover codes crosswalked to the 3 agriculture groups used by SJRWMD for model parameterization.

About half of data users indicated they need level 3 classifications at least sometimes and half indicated that level 3 uses are always needed (Table 3 and Figure 6). The following table (Table 3) summarizes user input for Level 2, 3, and 4 classification for each of the 9 Level 1 land uses. It is notable from Figure

6 that users most commonly indicated they always need Level 3 LU descriptions for Agriculture (2000 code LU uses), as many data users are also using FDACs’ FSAID data, level 2000 LU might be a sensible opportunity to reduce LULC update efforts.