Policy-Driven Urban Expansion and Land Use/Land Cover Change in Ewa, Honolulu (2002–2022): Remote Sensing and Machine Learning Analysis of Transit-Oriented Development Impacts

Abstract

While prior studies on LULC change in the Ewa region of O’ahu Hawai’i have explored the policy implications and the rapid infrastructure changes on land use, very few studies have attempted to fully integrate both of these changes in a comprehensive, long-term study of island geographies. Most of the past work has focused on general trends or short-term fluctuations, without considering the play of nuanced interactions between urbanization policies, transit-oriented development, and constraints of Hawai’i’s finite land resources. To fill these gaps, this study examines LULC changes in Ewa, Honolulu between 2002 and 2022, which emphasizes the impacts of strategic urban policies and infrastructure development, such as the Honolulu Skyline Rail Transit System. Using Landsat 7 satellite imagery and random forest machine learning classifier, in Google Earth Engine, LULC is classified into urban, forest, vegetation, barren, and water with classification accuracy of over 85%. The results highlight trends of significant urban growth especially after 2010, and highlight key issues of tension between housing demands and environmental sustainability in O’ahu. This study highlights the potential of integrated remote sensing and policy analysis for informing sustainable development in land-constrained island settings, and advocates for planning frameworks that more effectively balance growth, ecosystem stewardship, and community welfare.

1. Introduction

O’ahu, the most populated island in the Hawai’ian Islands, experiences urgent land use and environmental challenges associated with rapid urbanization and population growth and policy-driven infrastructure expansion [1,2,3]. The transformation of agricultural and natural landscapes to urban construction has increased the worries of sustainability, access to housing, and environmental equity in the region [4,5]. Given the island’s geographic limitations and unique ecosystems, such modifications have long-term consequences for the people and the policymakers [6].

Despite the importance of these trends, there are very few data-driven analyses that link policy efforts with spatial patterns of urban growth [7]. Previous studies have tended to be narrow in their focus (on one aspect of land cover loss or aspects of urban planning) and have not systematically connected policy interventions, TOD, and landscape transformation [8]. There is an urgent need to understand how specific policy tools and large-scale infrastructure works together to reorganize land use and social, economic, and ecological trade-offs [9].

The growth of tourism fueled the development of resorts and infrastructure, further impacting land use and coastal areas [5]. These changes have had profound effects on O’ahu’s environment and society, including habitat loss [1], increased pollution [10], and social and economic disparities [4]. Understanding these historical trends is essential for comprehending the current state of O’ahu’s landscape and for planning future land use changes. The increasing pressure from urbanization in O’ahu, coupled with the decline of the sugarcane plantations, created strong pressure to open agricultural land for development [6]. O’ahu has strong legislation to protect prime agricultural land from being transformed into development land, requiring a complex rezoning process that the state senate must approve.

This study focuses on the Ewa region of O’ahu, where land use and land cover (LULC) have changed rapidly and significantly. For the purposes of this study, we define the Ewa region based on the Census County Division (district), which encompasses partial areas of three sustainable development plan areas: Ewa, Central O’ahu, and Urban Honolulu (See Section 3.2 below). Population growth, strategic infrastructure projects—such as the Honolulu Skyline Rail Transit Corridor—and the region’s proximity to Honolulu’s urban core have all accelerated this transformation. Located between the Wai’anae and Ko’olau mountain ranges, Ewa exemplifies how urbanization, environmental pressures, and socioeconomic forces intersect. Its unique position makes it an ideal case for understanding how LULC changes unfold within the constraints of an island ecosystem.

Urbanization in O’ahu, particularly in the Ewa region, presents complex challenges that underscore the importance of this research in academic and practical realms. Since 2000, the region has experienced significant changes driven by population growth, infrastructure development, and policy initiatives, including the construction of Honolulu’s Skyline Rail Transit System. These urbanization trends have led to pressing issues such as the loss of agricultural land, high housing demand amidst constrained geographic resources, and mounting environmental pressures.

Land use and land cover changes provide a valuable framework for examining the intricate relationship between human activity and environmental transformations, particularly in regions experiencing rapid urbanization. Hawai’i’s finite land resources, unique ecosystems, and socioeconomic evolution make these changes especially significant. This situation is emblematic of broader patterns across Small Island Developing States (SIDS), where limited land, ecological fragility, and pressure from external economic forces create similar development challenges [11]. On the island of O’ahu, the Ewa region—a diverse area of 105,408 acres nestled between the Wai’anae and Ko’olau mountain ranges—has emerged as a focal point for urban expansion. Its strategic location underscores the need to explore how urban development, land preservation, and sustainability intersect.

This study uses advanced remote sensing techniques to analyze LULC transformations in Ewa from 2002 to 2022, offering insights into Hawai’i’s constrained yet dynamic land use patterns. LULC encompasses land functions such as agriculture, housing, or conservation and physical surface types like sand, vegetation, or water. Driven by urban expansion, population growth, and agricultural intensification, these changes reflect the ongoing tension between development and environmental stewardship.

This research helps to close that gap by using state-of-the-art remote sensing and machine learning methods to map 20 years of land use and land cover change in Ewa, O’ahu [12,13]. By combining Landsat satellite imagery with powerful classification algorithms in a cloud-computing environment, the study measures the impact of key urban policies—including the Honolulu Skyline Rail Transit System—on urban growth and resource distribution [14]. This methodological framework can offer a new approach to visualize, analyze, and understand the result of policy decisions within geographically constrained territories that are exposed to environmental pressure [15].

The findings reveal how policies, from transit-oriented development strategies to housing affordability reforms, have shaped these transformations. By capturing spatial and temporal patterns in the Ewa region, this study offers practical insights that bridge academic research and urban planning. It also introduces forward-looking strategies such as form-based zoning, which emphasizes the form and function of built environments, and effect-based planning, which evaluates development through environmental and social outcomes.

The novelty of this research is that it is considered a holistic approach, where a combination of geospatial analyses, assessment of urban policies, and subsequently the use of the results for future land planning is undertaken [16,17]. The analysis provides historical trends and generates actionable insights for sustainable urban development [18,19]. By focusing on integrated planning solutions and evidence-based decision-making, this study provides actionable tools and strategies on how to balance pressures of urbanization with ecological stewardship and community well-being [20,21].

Ultimately, these findings are meant to help decision-makers, planners, and scholars balance economic opportunity and environmental integrity in such a way that urban development in O’ahu grows in line with local needs and global sustainability issues [8] (Board of Water Supply, 2016).

This study is a land use and land cover (LULC) change analysis of the island of Ewa, Honolulu, between 2002 and 2022 based on Landsat 7 imagery and a random forest machine learning classifier. The analysis does show significant urban expansion, especially between 2010 and 2022, and that this was strongly driven by the Honolulu Skyline Rail Transit System and associated transit-oriented development (TOD) policies. While areas of the forest were relatively stable, this urban growth took place largely at the expense of barren and non-forest green lands, highlighting the continuing tension between the need for housing and the sustainability of the environment. Hence, the research underscores the influence of infrastructure and policy on land use patterns and advocates for comprehensive approaches to balancing urban development with ecological conservation.

2. Literature Review

Urban sprawl refers to the expansion of urban areas into previously rural or undeveloped regions, significantly impacting land use and environmental conditions [7]. Housing developments often transform agricultural and forested landscapes into residential zones [8]. Since the 2000s, shifts in demographics, economic trends, and housing preferences have reshaped urban expansion patterns [9]. Millennials, in particular, tend to delay purchasing homes [22] and favor urban or pedestrian-friendly suburban areas with access to amenities [19]. This trend has fueled demand for mixed-use developments that integrate housing, commercial spaces, and recreational areas in walkable environments [23].

The COVID-19 pandemic accelerated remote work, increasing demand for housing in exurban and rural regions [24]. Rising property costs in urban centers have further driven people toward suburban developments, intensifying urban sprawl [25]. Mixed-use suburban areas reduce reliance on vehicles by integrating workplaces, residences, and amenities nearby [26]. These developments tend to lower housing prices while increasing rental costs [27]. While single-family homes remain popular, suburban areas are witnessing a growing interest in multi-family housing, reshaping traditional development models [28].

Migration profoundly influences land use and urbanization. Historically, rural-to-urban migration propelled urban growth, often at the expense of agricultural land and ecological health [29,30]. Recently, counter-urbanization—where individuals move from cities to rural areas—has gained traction in affluent nations. While this shift can stimulate rural economies and preserve cultural landscapes [31], it also creates infrastructure strains, raises property values, and displaces long-term residents [32,33]. Additionally, speculative land investment near urban centers escalates rural land prices, accelerating unsustainable development [34] and exacerbating environmental degradation and housing inequities [35].

Urban expansion presents significant environmental risks, including deforestation, biodiversity loss, and increased flood susceptibility due to impermeable surfaces [36,37]. Reconciling economic development with ecological preservation remains a critical challenge [37,38]. Competing demands for agriculture, recreation, energy production, and tourism necessitate strategic planning tools such as zoning regulations and conservation easements [34,39]. Sustainable tourism approaches can help balance economic growth with rural preservation, ensuring long-term ecological resilience [40].

2.1. Urbanization in Hawai’i

The global patterns of urbanization and competition for land development are magnified dramatically within island settings of Hawai’i. Island environments, which are geographically isolated, have limited land resources, and contain fragile ecosystems, are important case studies for the extreme pressures of modern development. The nature of housing shortages, conversion of agricultural land, and environmental degradation are not just here, but are compounded, leaving sustainable planning as a survival and resilience imperative. Thus, land use change in Hawai’i provides a singular window through which to view the increased tensions between growth and conservation, with lessons that have particular, and highly applicable, resonance for other land-constrained areas of the globe.

Urbanization in the United States has significantly impacted rural land use, particularly in Hawai’i, where geographic constraints and demographic shifts present unique challenges. Rapid urban growth has led to the transformation of agricultural and forested lands into urban zones to accommodate economic expansion and population increases [7,30]. On O’ahu, the island’s limited land area amplifies these pressures, raising concerns about balancing development with agricultural and environmental preservation.

The struggle to reconcile urban expansion with farmland protection is evident in O’ahu’s Ewa Development Plan, initially introduced in 1997 and updated in 2013. Ewa was designated a key growth center, with its population projected to rise from 68,700 in 2000 to 164,000 by 2035 [41]. Between 1970 and 2019, urban land expanded by 18,572 acres, primarily in Ewa and Central O’ahu, at the cost of 18,137 acres of farmland [42]. The conversion of pristine agricultural land to urban land is also a major concern across the Caribbean, where islands such as Trinidad and Tobago are witnessing rapid yet low-density urbanization at the expense of their agricultural sector [43].

Hawai’i’s strict zoning laws complicate housing accessibility, restricting residential use to 4% of state land and allowing multi-family dwellings on only 0.3% [44]. Rising housing costs have prompted legislative efforts, including a 2024 policy increasing residential density and the 2025–2028 Strategic Housing Plan promoting affordable housing.

Despite these initiatives, development and conservation remain in tension. Projects like the Honolulu Rail aim to curb sprawl but often convert agricultural land. Broader strategies such as the Aloha+ Challenge 2014 and Hawai’i 2050 Sustainability Plan focus on smart growth and environmental stewardship to ensure long-term sustainability.

2.2. Sustainability in Hawai’i

Cultural preservation remains foundational. The principle of aloha ‘āina ensures that Native Hawai’ian values and traditional land stewardship practices inform modern planning, yielding place-based solutions that align environmental, social, and cultural sustainability goals [19].

Hawai’i Island, the largest and most ecologically diverse in the region, faces the challenge of balancing urban expansion with sustainable development. As its population and economy grow, strategic land use planning is essential to preserve natural resources and shape responsible development. A key approach is smart growth, which encourages compact, mixed-use communities while minimizing urban sprawl and protecting agricultural and open spaces. The Hawai’i 2050 Sustainability Plan emphasizes development in designated growth areas to safeguard ecosystems and maintain rural character [19].

Simultaneously, the state’s goal of achieving 100% renewable energy by 2045 necessitates strategic placement of solar, wind, and other renewable energy infrastructure. Urban projects that incorporate energy-efficient designs, microgrids, and renewables enhance resilience and advance long-term energy security [21].

Water resource management is equally essential. Watershed protection and the use of low-impact, green infrastructure help secure urban water supplies [45]. With climate change intensifying sea level rise and severe storms, adaptive planning that incorporates these risks into development decisions is increasingly necessary [46].

Finally, transportation plays a major role in sustainability. “Complete Streets” policies promote walking, biking, and public transit, reducing emissions and fostering healthier communities [47]. Transit-oriented development near hubs creates compact, pedestrian-friendly neighborhoods.

2.3. Transit-Oriented Development in Hawai’i

Transit-oriented development (TOD) is a cornerstone of contemporary urban planning, especially in land-constrained regions like O’ahu. In Honolulu, the development of a dedicated rail transit system illustrates how TOD can reshape urban landscapes by encouraging compact, mixed-use neighborhoods aligned with public transit. Although discussions about light rail in Hawai’i began long ago, the Honolulu City Council formally approved a fixed guideway system in December 2006 to enhance mobility and steer urban development along critical transit corridors [48]. This initiative advanced with a statewide ballot measure in 2008, which endorsed a “steel wheel on steel rail transit system” [49].

TOD, as coined by Peter Calthorpe [50] in his seminal work The Next American Metropolis [14], promotes walkable, mixed-use communities with diverse transportation options—from rapid transit and cycling to pedestrian walkways. In O’ahu, TOD reduces greenhouse gas emissions and urban sprawl and maximizes limited land resources while lowering infrastructure costs.

The evolution of O’ahu’s rail transit plan, culminating in the Honolulu Skyline Rail System, reflects decades of political, legislative, and urban planning efforts. Bipartisan support from key city leaders like Mayors Mufi Hannemann (2005–2010) and Kirk Caldwell (2013–2021) helped secure funding through local tax initiatives and a USD 1.55 billion grant from the Federal Transit Administration. Key legislation such as Bill 46 (2010) authorized a General Excise Tax (GET) surcharge for rail funding, while Senate Bill 4 (2017) extended the surcharge through 2030 and added USD 2.4 billion in support [51].

Despite these advances, the rail project has faced legal disputes over environmental reviews, construction delays, and cost overruns—challenges common to large-scale infrastructure in geographically limited areas. To support TOD implementation, Honolulu passed Ordinance 19-8 (2019), which mandates mixed-use development within 0.5 miles of transit stations. The ordinance encourages taller buildings, dedicates 10–20% of new developments to public space, and estimates a saving of over 50,000 daily commute hours [52].

Further reinforcing sustainable growth, the City and County of Honolulu adopted the Primary Urban Center Development Plan Update in 2023, following the enactment of Bill 24 in 2024. This updated framework addresses long-term challenges like climate resilience and housing affordability. Notably, nine neighborhood-level TOD plans—four extending into the Ewa study region—promise to concentrate urban development while preserving essential agricultural and conservation lands along the Skyline corridor.

2.4. Planning Concepts

Form-based zoning and effect-based planning represent two distinct yet complementary ways to address the multifaceted challenges of urban development. Form-based zoning prioritizes the physical appearance and spatial configuration of urban spaces rather than traditional land use classifications. By focusing on the appearance, scale, and relationship of buildings and public spaces, it fosters walkable, cohesive, and aesthetically pleasing neighborhoods [53,54]. This model encourages mixed-use development and better integration of built environments, shifting the focus from separating land uses to shaping the overall urban form.

In contrast, effect-based planning—or performance zoning—prioritizes the outcomes of development rather than its physical form. It sets performance standards related to environmental impacts, social equity, and economic viability [16,17], offering flexibility while ensuring the development aligns with broader community goals. By evaluating the consequences of urban growth, this approach supports adaptive, outcome-driven planning.

Together, these strategies can facilitate innovative, sustainable solutions in fast-growing areas like Ewa, where balancing rapid urbanization with sustainable development and community well-being is critical.

2.5. Spatial Analysis

The rapid expansion of Central O’ahu and Ewa underscores the urgency for a comprehensive, data-driven approach to land use analysis in Hawai’i. This strategy must effectively detect and assess land cover changes while integrating these shifts with broader policy trends and demographic patterns. Such insights are critical for sustainable development and informed policymaking in a region characterized by geographic limitations and environmental sensitivity.

Google Earth Engine (GEE), equipped with long-term Landsat imagery and machine learning techniques such as the random forest (RF) classification algorithm, provides a powerful analytical framework [13]. Landsat 7, with its 30 m spatial resolution and extensive temporal coverage, supports long-term monitoring of spatiotemporal trends. Meanwhile, GEE’s cloud-computing capabilities enable rapid processing of large-scale satellite datasets [15,55]. The RF algorithm’s ability to manage complex spectral relationships ensures accurate land cover classification and facilitates consistent, repeatable analysis [56].

This advanced approach aligns with Hawai’i’s commitment to evidence-based policymaking, exemplified by initiatives such as the Aloha+ Challenge Dashboard of 2014. It offers valuable insights into pressing issues like housing demand, resource conservation, and climate resilience.

Since 2000, land use changes in Central O’ahu and Ewa reflect a delicate balance between urban growth, agricultural preservation, and environmental protection [20]. Ewa’s rapid development, driven by projects like Kapolei, contrasts with evolving policies that emphasize sustainability and affordability [57]. Future efforts must refine these methodologies to support responsible development while safeguarding Hawai’i’s distinct agricultural and natural resources [18].

Given Hawai’i’s limited land availability, ongoing urbanization, and strong sustainability focus, an advanced technological approach to land use analysis is imperative [58]. The integration of Google Earth Engine, Landsat 7 data, and random forest classification offers a robust solution, equipping policymakers with crucial insights to balance growth, conservation, and community well-being across Hawai’i’s rural and urban landscapes [12].

In recent years, urban sprawl and land use dynamics and transit-oriented development (TOD) have attracted widespread scholarly interest worldwide and in Hawai’i, especially with respect to changing agricultural, forested, and rural landscapes as a result of housing development [7,8]. Drawing on recent research by Dieleman [23] and Alig et al. [1] that analytically link demographic change and migration patterns to more shifting urban form, this paper argues that the delayed home-buying of Millennials and post-pandemic remote work have fueled mixed-use suburban growth [19,22]. The case of Hawai’i exemplifies these wider trends, but geographic constraints and stringent zoning laws lead to some unusual outcomes: only 4% of land in the state is zoned for residential use and only on a total of 0.3% are multi-family units allowed [47].

While these papers provide valuable contributions to our understanding of land conversion and trade-offs between policy, they do not provide systematic mapping of the spatial impacts of policy and infrastructure interventions in Hawai’i in commonsense time. For example, prior studies have identified conflicts between rapid urbanization and agricultural land preservation [41,42], but there has been no quantified, spatially based evidence of how transit infrastructure such as the Skyline Rail speeds up urbanization along transit corridors or impacts open spaces. Further, research focused on smart growth and sustainability [19,41] emphasize place-based environmental and cultural stewardship but seldom combine these priorities with powerful geospatial analysis and machine learning methods.

Therefore, the present study represents an important methodological contribution with the use of Landsat 7 satellite images with Google Earth Engine and random forest classification. The latter makes it possible to measure land use/land cover changes with high accuracy over multi-year intervals [12,13,15]. Unlike previous studies, which were either generalizing land conversion trends or looking at relatively short periods, this study directly estimates spatial impacts and attributes them to key regional policies—the Ewa Development Plan, TOD legislation, and recent affordable housing programs [20,59].

Despite these successes, problems still exist. The studies performed to date still overlook the interconnected nature of housing affordability, migration and environmental risk in the island context [37,38]. Second, few past analyses have included local knowledge and cultural value systems, including aloha ‘āina [19], in a scientifically rigorous and practically actionable way.

3. Methodology

3.1. Background to Ewa Region, O’ahu

As the largest city in O’ahu, researchers projected the Honolulu metropolitan area’s population to reach 928,000 by February 2025, with an estimated annual growth rate of 0.87%. However, data from the U.S. Census Bureau show a 2.7% decline between 2020 and 2023, with the population dropping from 1,016,507 to 989,408. This apparent contradiction stems from multiple factors, including the effects of the COVID-19 pandemic and shifting migration patterns. While long-term projections anticipate gradual growth, actual rates may fluctuate in response to economic, social, and environmental influences.

Table 1. Population of Ewa Census County Division (CCD), City and County of Honolulu, and the State of Hawai’i.

Year (Decennial)	Ewa CCD	Honolulu	Hawai’i
2020	360,841	1,016,508	1,455,271
2010 (Estimate)	314,730	917,907	1,317,421
2000	272,328	876,156	1,211,537

Source: US Census (Accessed in January 2025).

provides a closer look at Hawai’i’s population.

A closer look at O’ahu reveals additional subtleties. Between 2010 and 2022, the island’s population grew by 4.1%, from 956,320 to 995,638, slightly lower than the overall population growth of Hawai’i (5.6%) and the United States (7.7%) during the same period. In recent years, between 1 July 2022, and 1 July 2023, the State of Hawai’i experienced a net population decline of 12 people per day. Although natural population growth added six individuals per day, net outmigration offset this increase, with 18 more people leaving the state than arriving each day.

Over the past two decades, the Ewa region has emerged as a focal point of urban expansion on O’ahu. According to the U.S. Census Bureau, Ewa’s population increased from approximately 272,328 in 2000 to 314,730 in 2010, reaching around 360,841 by 2020—a 33% increase. This growth reflects several driving forces, including economic opportunities, enhanced infrastructure like the Honolulu Skyline Rail Transit System, and the region’s strategic location between the Wai’anae and Ko’olau mountain ranges.

While this population surge signals Ewa’s regional vitality, it has also exacerbated challenges related to housing affordability, public service capacity, and competition for limited agricultural and conservation lands. These demographic trends offer critical insights for urban planners and policymakers. They highlight the need for sustainable land use strategies—such as form-based zoning and effect-oriented planning—that can manage growth while preserving Hawai’i’s environmental integrity and cultural resources [60].

3.2. Study Area

We selected the study area in the Ewa region based on several factors: (i) higher rate of population increase in the region compared to other regions, (ii) decision to make Kapolei, which is within the Ewa region, as the second city, and (iii) Segment 1 of Honolulu’s Rapid Rail Transit System, i.e., the Skyline, which operates from Kualaka’i (East Kapolei) Station to Hālawa (Aloha Stadium) Station. This area spans three of the Cities and Counties of Honolulu’s sustainable development plan zones: Ewa, Central O’ahu, and the Primary Urban Area.

As we are interested in areas closer to the Skyline, we chose our study boundary using the 2020 Census County Division Boundary labeled ‘Ewa,’ available through the Hawai’i Statewide GIS Program website. Although this boundary is smaller than the combined area of the three sustainable development zones, it fully encompasses all stations within Segment 1 of the Skyline system. Figure 1 illustrates this study’s area of research.

We used supervised classification in GEE to label training data and classify satellite images into different LULC classes. The following steps outline our supervised classification process.

3.3. Data Acquisition and Data Preparation

Our study used a collection of surface reflectance images from Landsat 7, which employs the Enhanced Thematic Mapper Plus (ETM+) sensor. This dataset, spanning from January 1999 to April 2022, included three key time points—2002, 2010, and 2022—selected for our analysis.

A known deficiency of Landsat 7 imagery might be the failure of its Scan Line Corrector (SLC) in May 2003, which results in wedge-shaped strips of missing data in images acquired after that time. This impacts our 2010 and 2022 analyses, and also images from the second half of 2002. We solved this problem pretty well in our image compositing method in Google Earth Engine. By taking a median composite of all images for a given year (e.g., 30 images for 2010 and 23 for 2022), a single final image without gaps was created for each study period. The approach is to take the median pixel value from all valid observations in the image stack; as the location of the data gaps change with each satellite pass, the median function will effectively fill in the missing data, so that the resulting image is spatially complete for use in classification.

The Landsat 7 images have a temporal resolution of 16 days, and an 8-bit radiometric resolution. We accessed and analyzed the USGS Landsat 7 Level 2 Collection 2 Tier 1 data through GEE. This collection includes atmospherically corrected surface reflectance (SR) with the following spectral bands:

Four bands in the visible and near-infrared (VNIR) range, two shortwave infrared (SWIR) bands, and one thermal infrared (TIR) band, which the system processes to retrieve surface temperature.

These bands are ortho-rectified, meaning they go through a geometric correction to eliminate distortions caused by factors like terrain and sensor angle, ensuring accurate geospatial alignment with the Earth’s surface [61]. For this study, we used six bands with characteristics summarized in

Table 2. List of original bands used for land use–land cover analysis.

Name	Wavelength (μm)	Description
SR_B1	0.45–0.52	Band 1 (blue) surface reflectance
SR_B2	0.52–0.60	Band 2 (green) surface reflectance
SR_B3	0.63–0.69	Band 3 (red) surface reflectance
SR_B4	0.77–0.90	Band 4 (near infrared) surface reflectance
SR_B5	1.55–1.75	Band 5 (shortwave infrared 1) surface reflectance
SR_B7	2.08–2.35	Band 7 (shortwave infrared 2) surface reflectance

Note. In all instances, min = 1, max = 65,455, scale = 2.75 × 10⁻⁵, offset = −0.2.

.

To enhance image quality, we applied pre-processing techniques that removed cloud and shadow interference, using the QA_PIXEL band, generated from the CFMASK algorithm, to mask out cloud (bit 3) and shadow (bit 4). We also filtered images using the CLOUD_COVER property, selecting only those with less than 30% cloud cover.

To improve classification accuracy, we incorporated several spectral indices commonly used in land cover analysis:

• Normalized Difference Vegetation Index (NDVI) to detect vegetation and forest areas using the formula [62]:

$$\mathrm{NDVI}={\displaystyle \frac{(NIR-Red)}{(NIR+Red)}} ,$$

• Built-Up Area Index (BUAI) to identify urban areas, using the formula [63]:

$$\mathrm{BUAI}={\displaystyle \frac{(SWIR1-NIR)}{(SWIR1+NIR)}}, \mathrm{and}$$

• Normalized Difference Water Index (NDWI) to delineate water bodies using the formula:

$$\mathrm{NDWI}={\displaystyle \frac{(G-SWIR1)}{(G+SWIR1)}}.$$

We also included surface slope data derived from the National Aeronautics and Space Administration’s (NASA) Shuttle Radar Topography Mission (SRTM) at a 30 m resolution [64] to refine classification.

Although Landsat sensors provide surface temperature data, we excluded it from our analysis due to its sensitivity to diurnal variation, atmospheric conditions, and surface emissivity, which reduces its reliability in distinguishing between our targeted land cover classes.

By integrating these indices and terrain, we develop a robust, accurate, and context-specific land use/land cover classification model tailored to the geographic and environmental conditions of our study area. We selected the years 2002, 2010, and 2022 for analysis to examine potential land use and land cover (LULC) changes under different political administrations and major infrastructure developments. The year 2002 marked the transition from Governor Ben Caetano (Democrat) to Governor Linda Lingle (Republican), representing a political shift in Hawai’i’s leadership. The second year, 2010, represented another transition, this time when Governor Linda Lingel handed over the position to Governor Neil Abercombie (Democrat). We selected these years to explore whether changes in political leadership between parties correlated with noticeable shifts in LULC patterns. Additionally, in 2010, voters approved the establishment of the Honolulu Authority for Rapid Transit (HART) as a semi-autonomous authority to build and operate the city’s rail system. Therefore, any land cover changes before 2010 were likely driven by state-level policies. Further, we chose 2022 because it marked the beginning of Governor Josh Green’s administration (Democrat) and coincided with the completion of Segment 1 of the Honolulu Skyline Rail, which opened to passengers in June 2023. Using these selected years, we created composite images for 2002, 2010, and 2022 (refer to

Table 3. Dates and number of images used to create composite images for 2002, 2010, and 2022.

Year	Dates Used	# of Images Used in Composite Images
2002	1 January 2002 to 1 January 2003	23
2010	1 January 2010 to 1 January 2010	30
2022	1 January 2022 to 1 January 2023	23

). After filtering low-quality images, we retained 23 images for 2002, 30 for 2010, and 23 for 2022.

To create a composite for each year, we applied the median function to the remaining images. This method calculates the median pixel value across all valid observations, producing a representative image for each respective year. This annual median composite approach was chosen specifically to normalize for seasonal variability, ensuring that the final images for each year were directly comparable for the temporal analysis of land cover change.

3.4. Training Data

To create the training dataset, we randomly assigned point features across five LULC classes—urban, forest, vegetation, barren, and water—for the years 2002, 2010, and 2022. We verified the accuracy of each point using high-resolution imagery from Google Earth Pro and satellite images available in GEE. We then confirmed the selected points against lower-resolution Landsat 7-SR composite images.

To ensure historical accuracy, we used November 2002 images to verify training points for 2002 and January 2011 images for 2010. For 2022, we used more recent 2024 satellite imagery available in GEE to cross-check the land cover features. While we could clearly identify urban and forest categories even in the Landsat 7-SR composites, we relied on the NDVI and BUAI to differentiate vegetation and barren categories more effectively.

Researchers in urban land use commonly use these LULC classifications. To ensure balanced representation across classes, we began with approximately 30 training points per category for the year 2002 and ran the RF classifier. Figure 2 illustrates the location of training points on the composite map. We visually assessed the classification results and added more training points as needed to improve accuracy. We continued this iterative process until the classification reached an estimated accuracy of over 85%. We repeated the same process for the years 2010 and 2022.

Table 4. Land use and land cover (LULC) classification description and number of training points in each class for the Years 2002, 2010, and 2022.

LULC Classes	Description	# of Training Points
		2002	2010	2022
Urban	Land cover as a result of any construction activities (e.g., buildings, roads, airports, solar farms, etc.)	205	200	231
Forest	Tree canopy with dense trees (e.g., forest, orchard, etc.)	180	172	178
Vegetation	Areas covering some vegetation and other than dense trees (e.g., agriculture farms, cropland, golf courses, urban parks, lightly covered bushes, etc.)	221	201	206
Barren	Areas where soil is exposed (e.g., barren lands, quarry sites, etc.)	152	151	157
Water	Water bodies (e.g., ponds, rivers, pools, etc.)	42	49	43
Total training and testing points		800	773	815

shows the LULC categories used in this study, along with the number of training points assigned to each class.

To determine the classification performance, a random partition was applied on the whole dataset of reference points. Eighty percent of the points were used to train the classifier and the other 20% were used for testing. By following this 80/20 split, the training and test datasets were completely independent of each other, and so the reported accuracy figures represent unbiased validation.

3.5. RF Classifier

In this study, we employed the RF algorithm within GEE as a robust tool for classifying LULC changes. Researchers widely adopt RF due to its consistently high classification accuracy, non-reliance on data distribution assumptions, and its ability to quantify variable importance [65]. This non-parametric method handles large datasets effectively—including those with outliers and noise [66,67]—by generating numerous decision trees from random subsets of training data and combining their outputs through majority voting. Although the internal split rules of its decision-making are not easily interpretable (i.e., it functions as a “black box”), RF reliably reduces overfitting and enhances generalization.

For our LULC classification, we implemented a supervised RF approach based on Breiman’s [68] methodology, which builds each tree using a bootstrap sample of the original training data. After testing various configurations, we selected 50 decision trees as the optimal number, offering a balance between model accuracy and computational efficiency [69]. To add detail, the implementation of the random forest classifier we used was from the Google Earth Engine (GEE). While the dataset was tested iteratively and the number of trees was fixed at 50, other hyperparameters (number of terms to split at each node) were left at their optimized default values as given by the GEE platform for land cover analysis. Increasing or decreasing this threshold did not yield significant accuracy improvements.

To assess classification performance, we conducted an accuracy assessment using two primary metrics: overall accuracy and the kappa coefficient as primary metrics [70]. Overall accuracy measures the proportion of correctly classified pixels relative to the total number of reference pixels across all classes. Mathematically, it is

$$\mathit{Overall} \mathit{accuracy} = {\displaystyle \frac{(TP+TN)}{(P+N)}}$$

where TP is true positive, TN is true negative, P is all positive, and N is all negative.

This metric measures the ratio of accurate classifications, and factors such as the number of land cover classes, spatial resolution, landscape complexity, and the quality of training and testing data can influence its value [71,72,73].

We further analyzed the classifier’s performance using a confusion matrix, which provided values for the producer’s accuracy (PA), user’s accuracy (UA), and the kappa coefficient, thereby offering a comprehensive evaluation of the RF classification model. The kappa coefficient is a statistical measure that evaluates the agreement between classified and reference data, accounting for the probability of chance agreement. We calculate the kappa coefficient as follows:

$$\mathit{Kappa} \mathit{coefficient} = {\displaystyle \frac{(\mathrm{P}\mathrm{o}-\mathrm{P}\mathrm{e})}{(1-\mathrm{P}\mathrm{e})}}$$

where Po is the relative observed agreement, and Pe is the hypothetical probability of chance agreement. Unlike simple percentage agreement, the kappa coefficient considers the probability of agreement between training and test datasets, providing a more robust reliability measure for categorical classification [74].

4. Results

4.1. Land Use and Land Cover Change Trends

After running the random forest-50 tree model for each of the three years, the results showed the area (in acres) and the percentage of coverage in each class (refer to

Table 5. Comparison of land use and land cover (LULC) classification and calculated area.

	2002		2010		2022
	Area (Acres)	%	Area (Acres)	%	Area (Acres)	%
Urban	22,486.55	21.41	23,227.87	22.17	30,393.91	29.15
Forest	46,455.74	44.24	45,220.21	43.16	45,961.53	44.08
Non-Forest Green	19,274.19	18.35	20,015.50	19.10	18,038.66	17.30
Barren	16,556.03	15.76	16,061.82	15.33	9389.99	9.00
Water	247.10	0.24	247.10	0.24	494.21	0.47

and

Table 6. Confusion Matrices for 2002, 2010, and 2022.

Panel A. 2002 Confusion Matrix.
	Urban	Forest	Non-Forest (Green)	Barren	Water	Total	User’s Accuracy
Urban	38	1	2	1	0	42	0.9048
Forest	1	31	3	0	0	35	0.8857
Non-forest (green)	0	0	45	1	0	46	0.9783
Barren	5	1	5	20	1	32	0.6250
Water	0	1	0	0	3	4	0.7500
Total	44	34	55	22	4	159
Producer’s Accuracy	0.8636	0.9118	0.8182	0.9091	0.7500		0.8616
Panel B. 2010 Confusion Matrix
	Urban	Forest	Non-Forest (Green)	Barren	Water	Total	User’s Accuracy
Urban	33	1	3	4	0	41	0.8049
Forest	0	37	1	0	0	38	0.9737
Non-forest (green)	1	1	41	0	0	43	0.9535
Barren	0	1	4	35	0	40	0.8750
Water	1	1	1	0	7	10	0.7000
Total	35	41	50	39	7	172
Producer’s Accuracy	0.9429	0.9024	0.8200	0.8974	1.0000		0.8895
Panel C. 2022 Confusion Matrix
	Urban	Forest	Non-Forest (Green)	Barren	Water	Total	User’s Accuracy
Urban	33	0	2	3	0	38	0.8684
Forest	0	39	1	0	0	40	0.9750
Non-forest (green)	2	0	24	2	0	28	0.8571
Barren	5	0	2	34	0	41	0.8293
Water	0	0	0	0	1	1	1.0000
Total	40	39	29	39	1	148
Producer’s Accuracy	0.8250	1.0000	0.8276	0.8718	1.0000		0.8851

Note. Panel A: training datasets = 640; testing datasets = 159; Panel B: training datasets = 599; testing datasets = 172; Panel C: training datasets = 667; testing datasets = 148.

, and Figure 3). The ranking of land cover types by area remained consistent across all three years, with the forest covering the highest percentage of land, followed by urban, non-forest green, barren, and water.

Between 2002 and 2010, the Ewa region of O’ahu experienced notable LULC changes (see Figure 4), driven by urban expansion resulting from population growth, economic development, and evolving land management policies. Urban areas expanded from 21.41% to 22.17% as planners and developers converted more land into built-up uses, such as residential, commercial, and infrastructural facilities. This conversion reflects a growing demand for housing and improved transportation infrastructure, which encouraged the redevelopment of previously undeveloped or lower-intensity land uses.

Simultaneously, non-forest green areas expanded from 18.35% to 19.10%, suggesting that transitional zones—often encompassing agricultural fields, suburban lawns, or low-intensity vegetative covers—continued to shift toward mixed-use developments. This trend reflects a shift in land utilization where landowners and developers increasingly adapt these semi-natural zones to support suburban growth and emerging economic activities.

Meanwhile, forest cover declined from 44.24% to 43.16%, highlighting the loss of natural vegetation. Selective clearance for urban and infrastructural expansion likely drove this reduction, even as conservation efforts attempted to mitigate extensive deforestation. Incremental forest loss became more evident as urban pressures intensified. Similarly, barren land decreased slightly from 15.76% to 15.33% as developers repurposed underutilized areas into urban or green spaces, reflecting efforts to rehabilitate land and increase land productivity.

Throughout this period, the water class remained constant at 0.24%, implying that water bodies naturally resisted change or benefited from strict environmental protections that limited alteration despite the rapid changes in surrounding land cover.

Overall, these LULC shifts from 2002 to 2010 illustrate a dynamic interplay between urban development and environmental conservation in the Ewa region, emphasizing the need for integrated strategies that balance growth with sustainable management.

Between 2010 and 2022, LULC changes continued to reshape this region. Urban areas grew substantially from 22.17% to 29.15%, while forest cover also increased, rising from 43.16% to 44.08%. In contrast, non-forest green areas declined from 19.10% to 17.30%, and barren land dropped from 15.33% to 9.00%, indicating a continued transformation of open and transitional zones into developed or vegetated land. However, the water area increased from 0.24% to 0.47%, further suggesting enhanced water management or expansion of water bodies through human intervention or natural processes.

4.2. Classification Accuracy Assessment

There is no universally accepted threshold for evaluating the accuracy of LULC classifications. Instead, researchers assess accuracy using multiple indices, including overall accuracy, the kappa coefficient, the user’s accuracy, and the producer’s accuracy [75,76].

In this study, the overall accuracy, which is the ratio of correctly classified pixels to the total number of pixels, exceeded 85% for all three classification years, with values of 0.861 (2002), 0.889 (2010), and 0.885 (2022). These results indicate a high level of classification reliability.

The kappa coefficient, which measures the agreement between the test data and validation data while accounting for chance agreement, also showed strong performance. The kappa values were 0.816 (2002), 0.857 (2010), and 0.847 (2022), indicating substantial agreement across all years (refer to

Table 7. Overall classification accuracy and kappa statistics.

Year	Overall Accuracy	Kappa Coefficient
2002	0.8616	0.8158
2010	0.8895	0.8567
2022	0.8851	0.8466

). Studies such as Mutale et al. [77] suggest a correlation between kappa and overall accuracy. The kappa and overall accuracy values in this study surpass the 85% threshold recommended by Congalton and Green [78], confirming that the maximum likelihood classification (MLC) method successfully identified the LULC categories.

The user’s accuracy, which represents the ratio of the number of correctly classified points for a particular class to the total number of points classified as that class, reflects how accurately the map represents features from the user’s perspective. In the 2002 classification, the user’s accuracy exceeded 85% for all categories except barren land and water. For instance, among 32 training points labeled as barren, the user correctly classified only 20 as barren. Similarly, in 2010 and 2022, the user’s accuracy remained above 85% for most categories, except for water (0.7) in 2010 and barren (0.829) in 2022. Due to the small sample size for the water class, the error margin was higher in those cases.

The producer’s accuracy, which is the ratio of the number of correctly classified reference points for a particular class to the total number of reference points for that class, evaluates the map’s reliability from the producer’s perspective. In 2002, all classes had producer’s accuracy above 85%, except non-forest green (0.818) and water (0.75). In 2010, only non-forest green (0.82) fell below this threshold. In 2022, non-forest green (0.8276) and urban (0.825) had slightly lower values, while all other categories surpassed 85%.

The reduced accuracy for the ‘Water’ and ‘Barren’ classes has been explained by the well-known problems with remote sensing classification. For the ‘Water’ class, the main reasons are its small footprint within the study area and consequently the smaller number of training samples, and the 30 m resolution of Landsat which can lead to ‘mixed pixels’ for smaller features such as streams or ponds. Similarly, for spectral confusion, some of the features in the ‘Barren’ class produced spectral confusion with similar features in the ‘Urban’ class, such as construction work sites or fallow agricultural fields, which could cause misclassification

4.3. Changes in LULC Classification

Among the five land use categories, forest and non-forest green areas experienced only slight changes from 2002 to 2010 and from 2010 to 2022. However, these changes are small in percentage and fall within the margin of error for LULC classification. As a result, we cannot conclude that forest or non-forest green areas underwent significant changes during this period. The data suggest that their land cover remained relatively stable between 2002 and 2022.

In contrast, urban land classification shows a significant increase in urban areas by 2022 compared to 2002 and 2010. This increase surpasses the classification error margin, confirming the growth as statistically meaningful. Specifically, the urban category expanded from 22.2% in 2010 to 29.1% in 2022. The construction and operation of the first segment of the Honolulu Skytrain largely drove this growth and accelerated urban development.

The urban expansion primarily drew from non-forest green areas, including agricultural lands and barren land. Notably, the barren category also experienced a significant decrease in the area from 2010 to 2022, further supporting the trend of urban conversion.

5. Discussion

5.1. Research Contributions

This study makes some important contributions to the understanding of the dynamics of LULC change in rapidly urbanizing island settings. First, it presents an integrated methodological framework that integrates remote sensing, machine learning classification, and policy analysis in order to investigate long-term spatial change. This methodology allows for a different, more nuanced understanding of the impact policy interventions (transit-oriented development (TOD) in this example) have on the patterns of urban growth.

Second, the research offers empirical evidence on the correlation of infrastructure planning and LULC changes, especially in the Ewa region. By comparing satellite imagery from both before and after TOD projects for 35 years, the study demonstrates the spatial and temporal nature of landscape changes brought on by TOD projects.

Third, the study adds to the general discourse on sustainable urban planning by bringing up the challenges and opportunities of development in land-constrained geographies. It highlights the need for balancing policy objectives with environmental realities, and illustrates how geospatial technologies can help inform decision-making based on data.

Finally, the study contributes to the academic debate by offering a link between technical analysis and policy assessment, thus providing a replicable framework for other regions dealing with the same urbanization pressures.

The changes in land use and land cover (LULC) from 2002 to 2022 reflect the combined influence of state government policies, urban growth patterns, and green development initiatives. These shifts are evident in a range of regional efforts, including forest protection, use of agricultural land, urban housing growth, installations of solar farms, and transit-oriented development (TOD). Strategic hubs, such as East Kapolei, UH West O’ahu, and Waipahu Transit Center (Pouhala Station), are pivotal in these efforts, demonstrating the region’s dedication to building integrated, sustainable, and transit-oriented communities.

5.2. Contextualizing Findings: A Comparison with Previous Research

A detailed comparison with the existing literature places our findings and explains their novelty. The observed increase in urban land cover in Ewa—from 21.4% to 29.1% in two decades—is consonant with global urbanization trends of other researchers such as Burchell et al. [15]. However, the nature and drivers of this change in a land-constrained island setting provides a unique perspective. For example, our findings are consistent with Bretschneider’s [10] analysis of agricultural land loss on the island of O’ahu, but extend it to include spatially explicit evidence, tying the agricultural land loss directly to the development corridor of the Skyline transit system.

The acceleration of urban growth after 2010, which coincided with the development of the rail project, represents a powerful real-world example of the transformative power of TOD as theorized by Calthorpe [50]. Our findings mimic other studies in other regions that empirically quantified the impact of new transit lines. For example, Zhang et al. [79] established that new metro lines in Shanghai significantly induced urban land conversion in a 1 km buffer of stations. Similarly, our analysis shows a high density of new urban development around the Skyline stations in Ewa. However, unlike the sprawling patterns we have often seen in a mainland context [80], the growth in Ewa is more intensely focused, bounded by geography and channeled by the linear path of the transit system. This highlights TOD’s potential as a growth management tool for geographically limited areas.

Methodologically, the approach of our study by applying the Google Earth Engine platform and the random forest classifier in a 20-year longitudinal study is in accordance with modern methods proposed by researchers such as Gorelick et al. [36] and Wang [81]. Many previous studies of land use in Hawai’i were based on aerial photography taken less frequently or required different classification algorithms that were more labor-intensive and less easily reproduced. Our framework shows a streamlined and robust way to go about this that can be used consistently for long-term monitoring, offering a major step forward for policy and planning in Hawai’i and other Pacific islands. By combining this state-of-the-art technical analysis with in-depth policy review, our work goes beyond the descriptive LULC mapping of many remote sensing studies to provide a more explanatory model of policy-driven landscape change.

5.3. Implications of the Results

5.3.1. Forest Conservation Land Use

The Ewa region has experienced notable changes in land use over the past two decades, particularly with respect to forest conservation (refer to Figure 5). State policies under different gubernatorial administrations have varied in their approach to balancing urban development with environmental preservation.

Governor Linda Lingle’s administration (2002–2010) introduced the Important Agricultural Lands (IAL) law, which, while primarily focused on preserving agriculture, indirectly supported forest conservation by limiting the conversion of forested areas into urban developments. Successive administrations built on this foundation and continued to emphasize the importance of preserving natural landscapes.

Governor David Ige’s administration (2014–2022) advanced these efforts by enacting Acts 152 and 153, which integrated land use planning with environmental policies, further reinforcing conservation efforts. As a result of these sustained initiatives, forest areas in the Ewa region have remained relatively stable from 2002 to 2022 despite increasing urban development pressures.

This long-term stability reflects the effectiveness of conservation-oriented policies and aligns with Hawai’i’s broader commitment to environmental protection, as outlined in the State Land Use Law, which reserves large portions of land for conservation purposes [82]. The consistency in forest cover observed over the study period suggests that successive gubernatorial administrations successfully upheld conservation priorities. Reforestation efforts and policies aimed at maintaining ecological balance have minimized forest loss, reflecting the region’s dedication to integrating land use and environmental policies.

5.3.2. Agricultural Land Use

Agricultural land use in the Ewa region has faced significant challenges due to the pressure for urban development. The decline of sugarcane plantations created opportunities to repurpose agricultural lands, but much of this land has gradually shifted toward urban uses.

Governor Neil Abercrombie’s administration (2010–2014) marked a shift towards sustainable development practices, with policies aimed at supporting local agriculture while accommodating urban expansion. As part of this strategy, the Ewa Development Plan established a Community Growth Boundary, which successfully protected 3000 acres of agricultural land. However, despite these efforts, the overall trend reveals a reduction in agricultural acreage (see Figure 6). The observed decline in non-forest green areas between 2010 and 2022 supports this, pointing to the conversion of agricultural lands into urban uses.

This trend reflects a broader challenge across Hawai’i—balancing agricultural preservation with the urgent need for housing. Governor Linda Lingle’s Important Agricultural Lands (IAL) law aimed to protect prime farmland. Yet, the ongoing housing crisis has continually pressured policymakers to rezone agricultural lands for residential development [83]. Governor Josh Green’s recent Emergency Proclamation on Housing underscores the ongoing tension between preserving farmland and meeting demand for affordable housing. The decline in non-forest green areas identified in this study likely reflects these conversions, underscoring the complexities involved in maintaining sustainable agricultural practices amid growing urbanization pressures.

Preserving agricultural land is a key component of sustainable urbanization in Honolulu. Conservation efforts aim to prevent urban sprawl from overtaking farmland, sustain local food production, and uphold traditional farming practices. The Community Growth Boundary in the Ewa Development Plan plays a key role in managing this balance, clearly delineating areas for urban growth and those reserved for agricultural use.

Beyond conservation, the region is increasingly embracing innovative approaches to urban agriculture. Community gardens, vertical farms, and the integration of farming within urban settings are becoming more common. These initiatives provide fresh produce to urban residents, promote food security, and reduce the carbon footprint associated with food transport.

5.3.3. Urban Changes: Housing Development

Housing development in Honolulu faces the dual challenge of accommodating a growing population while maintaining affordability and preventing urban sprawl. Rising housing demand has led policymakers to rezone agricultural lands for residential use. However, this strategy requires careful planning to balance the urgent need for housing with the long-term goal of preserving vital agricultural lands. Affordable housing initiatives remain essential to provide safe, stable living options for all residents, particularly Native Hawai’ans and low-income families.

In the Ewa region, urban change has been particularly pronounced, closely linked to housing development policies and the construction of Honolulu’s Skyline Rapid Rail Transit System. Between 2010 and 2022, urban land cover in Ewa expanded from approximately 22% to over 29%, coinciding with major phases of Skyline construction. This growth reflects policy decisions under Governors Linda Lingle, Neil Abercrombie, and David Ige, who championed transit-oriented development (TOD) as a model for sustainable urban growth (City and County of Honolulu Department of Planning and Permitting, 2013 [19]). Their administrations sought to address housing shortages while promoting compact, walkable communities integrated with public transportation.

Governor Josh Green’s 2023 Emergency Proclamation on Housing accelerated housing developments by temporarily suspending certain regulatory barriers to fast-track affordable housing construction [84]. While this proclamation sparked concerns about potential environmental impacts due to relaxed regulations, it underscores the urgency of addressing Hawai’i’s housing affordability crisis. The expansion of urban areas from 22.17% in 2010 to 29.15% in 2022 is largely due to the Honolulu Rail Project (Skyline) and related developments. Key TOD hubs such as East Kapolei, UH West O’ahu, and the Waipahu Transit Center exemplify efforts to integrate housing with public transit infrastructure and create mixed-use, pedestrian-friendly communities. Governor Green’s administration continues to push aggressive housing policies to boost the supply of affordable homes.

Table 8. Gubernatorial Land Use Trends in Hawai’i Since 2002.

Governor	Period (Years)	Key Land Use Policies	Legislative Actions	Source
Linda Lingle (R)	2002–2010	- Prioritized economic growth and urban expansion - Established Kapolei as O’ahu’s “Second City” - Enacted Important Agricultural Lands (IAL) law	- Hawai’i State Planning Act amendments to protect agricultural lands - Supported rail transit planning	State of Hawai’i (2005) [85,86]; Bretschneider (2023) [10]
Neil Abercrombie (D)	2010–2014	- Transitioned to sustainable development practices - Focused on transit-oriented development (TOD) for Honolulu Rail Project	- SB 4 (2011) to fund rail construction - Revised Ewa Development Plan (2013) with TOD incentives	State of Hawai’i (2012) [82]; City and County of Honolulu (2013) [19]
David Ige (D)	2014–2022	- Emphasized balance between development and conservation - Integrated land use with environmental policy	- Acts 152 and 153 (2021) merging land use/environmental roles - Extended GET surcharge for rail funding (SB 4, 2017)	State of Hawai’i (2021) [84]; Associated Press (2024) [5]
Josh Green (D)	2022–present	- Declared housing crisis an “existential” threat - Temporarily suspended regulations to accelerate housing development	- Emergency Proclamation on Housing (2023) - 2025–2028 Strategic Housing Plan, emphasizing TOD	Office of the Governor, State of Hawai’i (2023) [71]; Associated Press (2024) [6]

Note. D = Democrat; R = Republican.

summarizes gubernatorial land use trends in Hawai’i since 2002.

Over the past two decades, Hawai’i’s policy landscape has evolved significantly, shaped by shifts in governance and development priorities. Governor Linda Lingle (2002–2010), a Republican, emphasized economic growth, promoted the rezoning of agricultural land for development, and initiated plans for the Skyline rail project. Her Democratic successors, Abercrombie, Ige, and Green, redirected focus toward sustainability, embedding conservation goals into urban growth strategies through TOD and renewable energy initiatives.

Since 2010, land use policies have aligned closely with the expansion of the Skyline system, increasing urban density around key transit stations like East Kapolei. These efforts also included legislative initiatives to fund rail development. Despite protections under the Important Agricultural Lands law, Hawai’i lost 18,137 acres of agricultural land between 1970 and 2019, underscoring the persistent tension between development and environmental oversight, a challenge highlighted by Governor Green’s emergency housing measures.

Recent policies have also supported Hawai’i’s 2045 renewable energy goal by encouraging solar farm construction on barren lands and investing in climate-resilient infrastructure. These integrated strategies signal a growing commitment to harmonize urban development with long-term environmental and energy objectives.

5.3.4. Solar Farm

Energy sustainability is another critical aspect of urban development in Honolulu. Integrating renewable energy sources, such as solar and wind power, into urban infrastructure, is vital to meet the city’s growing energy demands in an environmentally responsible way. Solar farms and rooftop installations have become increasingly common, directly supporting Hawai’i’s goal of achieving 100% renewable energy by 2045. In addition, energy-efficient building designs and smart grid technologies further enhance the sustainability of the urban environment. Figure 7 illustrates the Ewa region’s land cover change due to solar farms from 2002 to 2022.

Hawai’i has prioritized the transition to renewable energy as a key component of its sustainable development agenda [87]. In the Ewa region, urban planning efforts have incorporated solar energy through policies that promote the development of solar farms on previously barren or underutilized lands. This approach supports the state’s 2045 renewable energy target and strategically reduces the carbon footprint of new developments.

Governor David Ige’s administration formalized this vision by setting the 100% renewable energy goal (Earth.Org, n.d.), and the expansion of solar farms in Ewa reflects the state’s commitment to that target. These solar projects serve as smart land use solutions, generating clean energy while preserving forested and agriculturally protected areas. By directing renewable energy infrastructure toward less ecologically sensitive land, the state supports urban growth without undermining conservation goals.

Governors Neil Abercrombie and David Ige have paved the way for Hawai’i’s renewable energy transition by prioritizing clean energy in policy and infrastructure planning. Governor Josh Green has continued and expanded these efforts, reinforcing the state’s long-term commitment to sustainable development. The continued emphasis on renewable energy in urban development ensures that Honolulu’s growth remains aligned with climate goals and environmental stewardship.

5.3.5. Skyline Transit Development

The Honolulu Rail Project (Skyline) has been a transformative force in shaping urban development in the Ewa region, significantly influencing regional land use patterns through transit-oriented development (TOD) principles. Key transit hubs such as East Kapolei, UH West O’ahu station, and Waipahu Transit Center exemplify sustainable urban planning in action.

The state is developing East Kapolei as a mixed-use village, combining residential units, commercial spaces, and community facilities like the Kroc Community Center. This integrated design promotes walkability and reduces dependence on automobiles [41]. The UH West O’ahu station serves as a gateway to the university and a catalyst for surrounding development by encouraging student-oriented businesses and diverse housing options. At the Waipahu Transit Center, planners have balanced historic preservation with high-intensity urban growth, showcasing how the Skyline project fosters sustainable urbanization while improving regional connectivity. Figure 8 exemplifies land use and land cover in East Kapolei as a result of Honolulu’s rail project.

As a key component of Honolulu’s broader urban strategy, the Skyline Rail System provides an efficient, environmentally friendly transportation alternative. By easing traffic congestion and reducing greenhouse gas emissions, it advances Hawai’i’s sustainability goals. TOD initiatives around transit hubs encourage the development of high-density, mixed-use, and walkable communities, decreasing reliance on personal vehicles and promoting a more sustainable urban lifestyle.

Changes in gubernatorial leadership have influenced the trajectory of large-scale projects like Skyline. For instance, fiscally conservative administrations may scrutinize project costs, while those prioritizing climate resilience may promote renewable energy integration and conservation-oriented land use. Investments in solar farms and forest conservation reflect Hawai’i’s broader commitment to clean energy and ecological preservation [88,89].

Governor Josh Green’s Emergency Proclamation Relating to Housing highlights the urgency of addressing the state’s housing crisis. By temporarily suspending regulatory barriers, the proclamation aims to fast-track affordable housing construction [83]. However, this approach has raised concerns about potential environmental impacts due to relaxed oversight.

Hawai’i continues to grapple with reconciling urban expansion, environmental conservation, and sustainable development. Successive administrations have sought to balance these competing priorities, from Governor Lingle’s focus on economic growth to Governor Green’s emphasis on housing affordability. Moving forward, innovative strategies that combine smart growth principles, renewable energy integration, and climate adaptation planning will be critical to ensure that urban expansion in Ewa aligns with the state’s sustainability goals.

From a planning perspective, the findings point to the necessity of forward-looking zoning and allocation of land that recognizes the wake-up effects of large transit projects. The concentrated, spatial growth patterns of urban development that occur around transit corridors indicate that TOD can be an effective tool for the concentration of development in an efficient way—if it is aided by complementary policies such as affordability, green space, and community.

Moreover, the study is an example of how geospatial analysis can be used to guide evidence-based decision-making. By combining satellite imagery with policy chronologies, stakeholders can better understand the impact of interventions in the past and tailor future interventions. This approach is especially valuable in areas such as Ewa, where land is scarce and the area is ecologically sensitive, requiring a balance between growth and conservation.

5.3.6. Urbanization Within Sustainability

The observed urban expansion has direct implications for the environment, starting with the loss of the fertile soil and the capacity to practice agriculture. Between 2010 and 2022, non-forest green areas (from 19.10 to 17.30%) and barren land (from 15.33 to 9.00%) decreased, largely to make way for the built environment. This shift to impervious urban surfaces permanently takes land potential for agricultural production, affecting local food security.

Furthermore, the highly large expansion of urban area changes the local water balance. The growth of impermeable surfaces can make areas more vulnerable to flooding, as well as surface runoff, which reduces the infiltration of rainwater that is necessary in order to recharge aquifers, and which can transport pollutions into coastal waters. Finally, although the construction process and the loss of carbon-sequestering vegetation has a carbon cost, a main objective of the transit-oriented development analyzed in this study is to minimize the long-term carbon footprint of the region by reducing greenhouse gas emissions from transportation.

Sustainable development is central to the planning of these junction points in the Ewa region. East Kapolei, envisioned as a mixed-use village, aims to reduce reliance on private vehicles by integrating residential, commercial, and transit facilities. This design reflects the core principles of transit-oriented development (TOD), which promote walkable, transit-accessible communities [17]. Concentrating housing development around these transit hubs helps address O’ahu’s housing shortage while minimizing urban sprawl. However, achieving this vision requires careful planning to balance increased housing density with the preservation of agricultural land and open spaces.

Urbanization in Honolulu presents a complex issue that planners must address through a sustainability-focused framework. Although Skyline intended to improve regional mobility, it also shaped urban development along the transit corridor. By encouraging higher-density development and lowering infrastructure costs, the project offers solutions to land scarcity. However, it also raises concerns about environmental degradation, the disruption of ecological systems, and the potential displacement of rural communities.

The continued expansion of urban areas has led to the conversion of agricultural lands and natural habitats, raising concerns about the loss of agricultural land and the need for stronger conservation efforts. The Ewa Development Plan identifies Ewa as a principal growth zone, which has led to significant changes in land use over time. While various initiatives have aimed to conserve agricultural fields and open spaces, the central challenge remains the following: how to balance the pressures of urban expansion with the long-term preservation of rural land and environmental resources. Similar tensions exist in other island economies heavily reliant on tourism, for example Fiji, where demand for resort-based infrastructure and coastal development is directly in conflict with traditional land uses and endangers the sensitive ecosystem [90].

Urbanization in Honolulu County, particularly in the Ewa region, must align with Hawai’i’s broader priorities of sustainability, cultural preservation, and community well-being. Ewa City’s development strategy reflects this alignment by implementing policies such as form-based zoning and transit-oriented development to balance growth with the protection of Native Hawai’ian values. At the heart of this approach is the principle of mālama ‘āina, or caring for the land, which the Ewa Development Plan reinforces through its Community Growth Boundary. This boundary protects over 3000 acres of agricultural land from urban encroachment, ensuring the preservation of vital spaces for local food production and the continuity of Native Hawai’ian farming practices [19]. These efforts conserve the natural environment and honor the deep cultural connection Native Hawai’ians maintain with the land.

The implementation of the Skyline Rapid Rail Transit System further supports sustainable urbanization by promoting TOD principles. Concentrating development around transit hubs in Ewa City fosters high-density, mixed-use neighborhoods that are walkable, well connected, and less dependent on private vehicles. This approach lowers greenhouse gas emissions and supports Hawai’i’s goal of achieving a carbon-neutral economy [41]. TOD also revitalizes neighborhoods by providing diverse housing options and stimulating new local employment opportunities.

Addressing housing affordability is another critical component of sustainable urbanization rooted in Hawai’ian values. Planners have prioritized affordable housing policies and the regulation of short-term vacation rentals to ensure that Native Hawai’ians and long-term residents can remain in their communities. This commitment to affordability reflects the Hawai’ian value of kuleana, which emphasizes the responsibility to care for others and foster equity in the community.

Alongside these efforts, urban development in Ewa City also emphasizes economic and social equity. Local initiatives increasingly focus on creating jobs and supporting small businesses in sectors like technology, renewable energy, and creative industries. These strategies aim to build a more self-sufficient economy and reduce dependence on tourism. Moreover, active community participation through public consultations, workshops, and collaborative planning remains central to the process, ensuring that development is inclusive, culturally sensitive, and aligned with residents’ aspirations [89].

By integrating sustainable urban planning strategies with core Hawai’ian values, Ewa City models a balanced and inclusive approach to urban growth. Through TOD, affordable housing, economic diversification, and active community engagement, Honolulu is paving the way for development that preserves its cultural identity while protecting its environmental heritage.

5.3.7. Resolution from Urban Planning Theories

Urban planning theories offer invaluable frameworks to address the challenges of rapid urbanization in Honolulu, particularly in Ewa City. Sustainable development in this context requires strategies that balance the pressing demands for housing with the protection of the environment and cultural heritage. Integrating robust planning paradigms—such as form-based zoning, effect-based planning, and transit-oriented development (TOD) informed by detailed land use and land cover (LULC) analyses can guide future development toward inclusive, resilient, and place-sensitive urban environments.

A critical aspect of this study’s findings is the difference between two patterns of urban growth: urban sprawl, and urban densification. Urban sprawl is the phenomenon of low-density development eating up previously rural and agricultural areas, often making us even more car-dependent and destroying open space. In contrast, urban densification is a process of population and building density growth in existing urban areas, often achieved by strategies such as building mixed-use developments and promoting taller buildings, and is one of the fundamental principles of form-based zoning.

Transit-oriented development (TOD) is, in its most basic sense, a policy tool to encourage conscious urban densification as a direct countermeasure to sprawl. The basic principle of TOD is to concentrate housing, commercial spaces, and amenities in walkable, mixed-use communities centered around high-capacity transit stations. By forcing growth into these nodes, TOD hopes to minimize the pressure of development on peripheral agricultural and conservation lands, accommodate the growth of the population sustainably, and maximize the use of public infrastructure.

In the Ewa region, massive urban growth during the period 2010–2022 (which happened to occur concurrently with the development of the Honolulu Skyline rail system) is indicative of this policy-driven density push, and not of uncontrolled sprawl. The analysis reveals that there was a heavy concentration of growth in and around designated TOD areas like East Kapolei as well as the Waipahu Transit Center that were specifically planned to be higher density, pedestrian-friendly communities. Therefore, although the results indicate a shift in land use from non-forest green and barren lands towards urban land use, the TOD framework supported this transformation in a way to achieve a compact and sustainable urban form. This controlled growth pattern, with emphasis on densification around transit, is in stark contrast to the unchecked, low-density growth that typifies urban sprawl.

Form-based zoning emphasizes the physical form, scale, and aesthetics of the built environment rather than segregating land uses. This planning approach promotes the development of walkable, mixed-use neighborhoods that blend residential, commercial, and recreational functions, contributing to vibrant, pedestrian-friendly corridors [17,54]. In Ewa City, LULC studies show that urban areas grew from 21.41% in 2002 to 29.15% in 2022, largely due to mixed-use developments near key transit nodes like the Honolulu Skyline rail stations at East Kapolei and UH West O’ahu. Such developments create visually appealing urban areas and convenient access to parks, shops, and community amenities.

Effect-based planning, or performance zoning, complements this approach by regulating outcomes rather than prescriptive design elements. This approach establishes performance benchmarks for environmental quality, social equity, and economic performance [16]. In Ewa, LULC data shows a decline in barren land from 15.76% in 2002 to 9.00% in 2022, suggesting the successful conversion of underutilized land into productive urban and green spaces. Effect-based planning supports this transformation by allowing flexibility in how developers meet performance goals, such as incorporating rain gardens, green roofs, and permeable surfaces to manage stormwater and mitigate heat. It also supports the development of diverse, affordable housing options tailored to Hawai’i’s multigenerational households and changing demographic needs.

Transit-oriented development (TOD) further reinforces sustainable urbanism by concentrating dense, mixed-use, walkable communities around public transit hubs. TOD reduces reliance on private vehicles, mitigates traffic congestion, and contributes to lower greenhouse gas emissions [41]. The Honolulu Skyline Rail Transit System has catalyzed TOD in Ewa, guiding development along the transit corridor. LULC analyses reveal that urban growth has concentrated around transit nodes, and non-forest green areas have declined, suggesting intentional efforts to direct development inward and protect agricultural and conservation lands on the periphery.

Central to these planning strategies is robust community participation. Engaging residents, especially Native Hawai’ans and long-term community members, helps align planning efforts with cultural priorities. The Hawai’ian concept of aloha ‘āina, or “love of the land,” underscores the importance of integrating traditional land stewardship practices into modern planning frameworks [89]. Planning processes that include public consultations, collaborative workshops, and co-design initiatives allow communities to shape their built environment in culturally responsive and equitable ways.

Despite considerable progress, planners still face the challenge of reconciling rapid urban growth with the preservation of Hawai’i’s natural and cultural landscapes. State policies have shifted across administrations—from Governor Lingle’s emphasis on economic growth to the sustainability-focused strategies of Governors Abercrombie, Ige, and now Governor Green, who has prioritized housing affordability through emergency legislation. These changes reflect an ongoing tension between meeting immediate housing needs without compromising long-term sustainable goals.

By integrating form-based zoning, effect-based planning, and TOD, anchored in comprehensive LULC analysis and strengthened by active community engagement, Ewa region can serve as a model for holistic urban development. This approach supports economic opportunity, cultural preservation, and environmental sustainability, offering a practical blueprint for managing growth in island and land-constrained regions. Indeed, the struggle to reconcile development pressures with sustainability is a defining challenge for planners in island regions, requiring integrated policies that protect finite coastal and agricultural lands from irreversible conversion, as documented in Barbados [91].

5.4. Benefits, Limitations, and Future Outlook

The key strength of this study is that the remote sensing-based 20-year longitudinal analysis is integrated within a comprehensive review of urban policy. Because it combines both policy and land cover data in a spatially explicit manner, this combined perspective offers a more nuanced explanation of how policy decisions manifest in the landscape than studies that focus on policy or land cover alone. Our methodology is transparent, cost-effective, and highly reproducible due to our use of publicly available Landsat data and the random forest classifier through Google Earth Engine for analyzing and identifying urbanization pressures in other areas. This framework can be a useful tool for evidence-based policymaking by enabling the empirical evaluation of large-scale infrastructure projects.

However, we recognize a few limitations. First, although the spatial resolution of the Landsat 7 imagery (30 m) is appropriate for regional scale analysis, it can make it difficult to accurately classify themes of fine-grained or heterogeneous urban character. This can cause small misclassifications especially in mixed-use areas where there are residential plots, vegetation, and small structures. Second, although our analysis indicates a strong relationship between the development of the Honolulu Skyline and accelerated urban growth, remote sensing data alone cannot provide a conclusive test of causation. Urbanization is a complex process based on various drivers such as market forces, economic tendencies, demographic changes, etc., not included in LULC analysis use directly. We attribute change in our study to policy, but these other factors are certainly involved.

Furthermore, this study is purely reliant on physical land cover and does not include socioeconomic data, such as shifts in population density, housing affordability, or community demographics. Such a compilation would present a more complete picture of the drivers and social outcomes of the observed urban growth. Finally, the random forest algorithm, although very accurate, has a “black box” nature which remains less interpretable in comparison with the other models for which it is possible to understand the spectral features driving the classification.

These limitations leave a number of fruitful directions for future research. For future studies, higher resolution satellite data (e.g., Sentinel-2 or commercial data) can be used to provide a more fine-scale analysis of urban morphology, overcoming the resolution limitations which have been studied here. A mixed-methods approach that combines quantitative socioeconomic data and qualitative data from stakeholder interviews would be invaluable in helping to better understand the causal link and determine the social equity implications of TOD. Finally, the historical LULC maps produced in this study can be used as a basis for the implementation of predictive models (e.g., Cellular Automata-Markov Chain models) to simulate urban growth scenarios in future time as a function of different policy and climate change conditions, offering a strong tool for proactive and sustainable urban planning.

6. Conclusions

Between 2002 and 2022, the Ewa region of Honolulu experienced significant changes in LULC, driven largely by rapid urbanization and the implementation of Honolulu’s Skyline Rapid Rail Transit System. This study used remote sensing data and the random forest classification technique to analyze the extent of Ewa’s urban expansion while examining its relationship to sustainable development practices and infrastructure demands.

Urban land use expanded noticeably over the two-decade period, becoming increasingly visible as urban development intensified. However, the most substantial leap in urbanization occurred after the launch of the Skyline project. Much of the land previously designated as agricultural or barren became a built environment, leading to a marked reduction in open space. This transition highlights the urban growth trajectory and the influence of varying preservation policies under different state administrations.

The study provides actionable insights to support more effective urban planning decisions. By applying advanced classification tools and geospatial analysis, it offers a data-driven foundation for integrating sustainable strategies into urban growth management. It also suggests planning approaches that reduce the unchecked spread of urban areas while supporting land transportation integration.

At the same time, the research accentuates the importance of balancing urban development with the conservation of natural resources. Policymakers must remain aware of the dual imperatives: to meet rising housing demands and to protect the state’s agricultural and ecological systems. This study advocates for a planning model that incorporates form-based and effect-based planning concepts to balance urban development and land conservation.

Future research could explore how transit-oriented development affects community engagement and cultural preservation in rapidly urbanizing regions. Additionally, studying the integration of renewable energy into urban planning could further enhance sustainability efforts. A fine-grained spatial accessibility analysis, using buffers around transit stations, could also provide a more granular understanding of how new infrastructure directly influences localized development patterns. Ultimately, this study reinforces the idea that ecological responsibility must stand alongside economic growth in land-limited regions like Honolulu.

By combining rigorous LULC analysis using machine learning methods with practical policy analysis, this study contributes to a deeper understanding of how urban development can coexist with environmental stewardship. The results offer strategies for achieving multiple objectives simultaneously, expanding housing, fostering economic development, protecting the environment, and preserving Honolulu’s rich cultural and ecological heritage.