1 Introduction

Plastics has become excessively prevalent on our planet. They have the potential to eventually reach the ocean by moving from land sources through waterborne, airbone or by direct deposit (Spring et al., 2021). Current estimates suggest that 70%–80% of plastic pollution originates from land-based sources, which contribute to 91% of oceanic plastic pollution (Lebreton & Andrady, 2019; Li et al., 2020). Substantial effort has been focused on direct and indirect observation of plastic debris in the ocean, with various studies demonstrating that the spectral signature of floating plastic debris can likely be characterized via remote sensing (Garaba et al., 2018; Hu, 2021; Papageorgiou et al., 2022; Sasaki et al., 2022; Topouzelis et al., 2019, 2020). Accurately identifying land-based sources of plastics is also essential for developing effective strategies to mitigate plastic influx into watersheds. However, mapping the extent of plastic on land globally has proven challenging. Attempts have been made to detect plastics in satellite images using visual pattern recognition Kruse et al. (2023), but this provides only suggestive visual evidence of dumping rather than definitive detection. Moreover, a focus on visually-distinctive waste dumping may overlook other sources of plastic pollution, such as exterior paints (Turner, 2021) and agricultural uses (Piehl et al., 2018), both of which shed plastic particles into the environment.

Fortunately, orbital imaging spectrometers have the capability to characterize the global distribution of surface plastic material (Schmidt et al., 2023; Zhou et al., 2022). Spectroscopic techniques can directly detect the diagnostic molecular fingerprints of some plastic polymers in reflected infrared light, adapting analytical laboratory techniques for remote measurement of the Earth surface's chemical composition. Here, we demonstrate the capability of a new NASA imaging spectrometer mission to map macroplastic signatures globally. Launched on 14 July 2022, the Earth Surface Mineral Dust Source Investigation (EMIT) is currently operational aboard the International Space Station (ISS), imaging latitudes ranging from +51.6$${}^{\circ}$$ to −51.6$${}^{\circ}$$.

EMIT acquires spectra spanning the whole solar-reflected range from 380 to 2500 nm with high sensitivity (D. R. Thompson et al., 2024). This range includes strong hydrocarbon absorption features in the shortwave infrared which can be used for plastic detection. EMIT's ground spatial sampling of $$60\times 60$$ m is best suited to detecting large targets, but matched filter detection methods are also capable of recognizing and quantifying subpixel concentrations by utilizing known spectral endmembers of the target materials.

Due to its primary mission to map the mineral composition of dust emitting areas of the globe, EMIT's first year and a half of coverage focused primarily on the arid regions, including the Western United States, Southwestern South America, North and West Africa, the Middle East, and Central and East Asia. We used EMIT observations over these regions from June to October of 2023 (Figure 2) to detect two types of plastics: High-Density-Polyethylene (HDPE) and Polyvinyl Chloride (PVC). These specific types of plastics were chosen for their widespread use in global agricultural sectors and their common presence in plastic waste (FAO, 2021; Wojnowska-Baryła et al., 2022). Additionally, they exhibit strong reflectance in the NIR wavelength bands covered by EMIT: 1215 , 1710, and 2313 nm.

2 Methodology

We focused on the detection of High-Density-Polyethylene (HDPE) and Polyvinyl Chloride (PVC) on the EMIT scenes. To map these spectral signatures within EMIT scenes, we first calibrated EMIT spectra to radiance units (D. R. Thompson et al., 2024) and then estimated the surface reflectance by inverting a radiative transfer model (D. R. Thompson et al., 2018). The EMIT reflectance products are available to the public in the EMIT Earthdata Search archive (Green, 2022). We excluded certain scene content that have the following criteria: clouds, recognized with multiple-channel thresholding of D. R. Thompson et al. (2014), augmented by a cirrus detection in a deep water vapor absorption channel (Gao et al., 1993); and surface water glints by flagging water pixels (Green, 2022). Finally, we analyzed the EMIT reflectance products with a matched filter detection algorithm (Manolakis et al., 2000), applied column-wise as in (D. Thompson et al., 2015). All EMIT scenes between June 1st and October 22nd of 2023 were analyzed.

The matched filter estimates the amount of additional plastic signal (target spectrum) in a measured spectrum beyond the background distribution. We selected a target spectrum to represent each category of plastic from the USGS spectral library version 7 (Kokaly et al., 2017). For PVC, we used USGS sample GDS338 and for HDPE we used GDS385. We limited the matched filter to the near- and short-wave infrared region (SWIR, 1100–2500 nm), which contains well-established diagnostic absorption features of plastics. Notable absorption centers within this range occur at approximately 1215 , 1417, 1537, 1732, 2046, and 2313 nm (Garaba et al., 2018). Wavelength bands associated with water vapor absorption within the ranges of 1300–1560 nm and 1770–2000 nm were excluded to mitigate potential residual contamination.

The matched filter returns a score that quantifies the similarity between the measured reflectance spectrum and a reference plastic spectrum. In this study, we use a detection threshold of 0.02 for PVC and 0.03 for HDPE. While a score above these thresholds suggests the presence of hydrocarbon-based materials, it does not imply a literal fractional coverage (e.g., a score of 0.02 does not mean plastic covers 2% of the pixel). Rather, the MF score means that the numerical projection of the measured reflectance onto the reference yields a magnitude of 0.02, relative to the unity value of a pure pixel. Several factors can lower the MF score beyond fractional coverage. These include spectral mismatches between the observed reflectance spectrum and the reference spectra, atmospheric scattering (which spreads the signal across pixels), and the optical properties of materials such as translucent agricultural plastics, which transmit light from underlying surfaces.

Given EMIT's spatial resolution of 60 m $${\times} $$ 60 m, plastic signals typically occur at subpixel scales, as large opaque plastic-covered areas are rare. In general, the simultaneous presence of absorption features near 1.7 and 2.3 $${\upmu }$$m is characteristic of hydrocarbon materials, since no other common Earth-surface materials exhibit both features. To validate these subpixel detections, we have checked for the presence of absorption depth of both features.

Additionally, we account for false positives by looking for spectral features of potential sources that overlap with some of the plastic features. For example, in scenes observed within the Sahara desert region, gypsum-induced false positives exhibit overlapping features with HDPE and PVC at 1700 nm but showcase a distinct feature at 2200 nm. Therefore, we have incorporated an automated condition to exclude spectra that exhibit spectral features at 2200 nm. For that, we have detrended the spectra by dividing the observed reflectance spectrum by a polynomial fit and then assessed the depth at 2200 nm. If the relative band depth was found to be negligible (i.e., $$\hspace*{.5em}$$ 1), it indicated the absence of a feature at 2200 nm.

To better illustrate the resolution of the EMIT data, we show in Figure 1 an EMIT RGB image of an area in Spain captured on 22 June 2023. in which we highlight the locations of pixels that meet our plastic detection criteria (MF score $${ >} $$0.02 for PVC or 0.03 for HDPE and presence of absorption features at 1700 and 2300 nm) using circular markers in the RGB image. The corresponding spectra are shown in the left panel and color-matched to the markers. Figures S4–S8 in Supporting Information S1 provide Google and Planet Scope Imageries for these marked locations.

3 Results and Discussion

Figure 2 displays the spatial distribution of detections of spectral signatures classified as HDPE or PVC. Circle sizes represent the density of MF detections within each grid cell. The map is constructed using a $$1000\times 1000$$ grid over latitude and longitude to count the number of HDPE and PVC inferences per cell. While detections are labeled as HDPE or PVC based on the highest MF score, this classification should be interpreted as indicative rather than definitive. The goal is to demonstrate that the MF approach can effectively flag pixels whose EMIT-measured spectra closely resemble known plastic signatures, particularly those of HDPE and PVC.

As shown, EMIT detections are primarily concentrated in populated regions, with PVC-classified detections notably clustered in Europe. HDPE exhibits a much broader distribution across our observations, with significant presence in Europe, the United States, and Asian countries such as Taiwan, South Korea, Japan, and the Philippines. While these detections are limited by the EMIT's latitudinal and longitudinal coverage, they provide valuable insights on the plastic distribution. Notably, we observe no detections in other regions covered by the EMIT, such as Southwestern South America and Australia, and almost no detection in the Northern Africa where coverage focus was largely on the Sahara Desert.

Upon evaluating individual detections, we find several cases associated with agricultural regions. Agriculture has relied extensively on plastic materials, with an estimated 12.5 million metric tons (Mt) of plastics utilized by this sector in 2019 (FAO, 2021). The demand for primary categories of land-based agricultural plastic materials, including greenhouse, mulching, and silage films, is projected to increase from 6.1 Mt in 2018 to 9.5 Mt by 2030 (Le Moine, 2018). An example of plastic detection associated with agriculture practices is shown in Figure 3a. In June 2023, EMIT detected spectral signatures indicative of HDPE, associated with a high concentration of greenhouses in Águilas, Murcia, Spain (Figure 3a, left). Google Street View imagery from May 2023, a month before EMIT overpass, consistently confirms the presence of these greenhouses. Another detection in June 2023 was observed in Sicily, Italy (Figure 3a, middle), with Google imagery from September 2023 (a month after EMIT overpass) also confirming the presence of greenhouses at that location. An additional detection associated with agricultural structures was observed in Hidalgo, Mexico (Figure 3a, right). Our findings underscore the potential of EMIT to accurately assess the widespread utilization of plastics in agricultural settings. A global understanding of these practices is important considering the intensification of the adoption of plastic materials in agriculture (Tong et al., 2024; Vox et al., 2016). The prevalence of plastic usage in this sector underscores its potential as a source of pollution if not adequately managed (UNEP, 2022). In Korea alone, it has been reported that over 50,000 tons of LDPE greenhouse film waste were generated annually from 2017 to 2020 (Kim & Lee, 2022). The breakdown of agricultural plastic can lead to microplastics polluting soils, and possible release of additives and chemicals. This, in turn, may affect soil characteristics, soil biota, plants, and adjacent water bodies (Briassoulis, 2023).

In Figure 3b we show an example of indication of HDPE in an aquavoltaic used as a renewable source. Due to Taiwan's limited and vulnerable land resources, the government has permitted the development of aquavoltaics on land designated for aquaculture (Hsiao et al., 2021). This site was not accessible via Google Street View, so the figure shows Google satellite imagery from 2025. Identifying SWIR plastic absorption features in aquaculture targets is challenging because these environments are dominated by liquid water surfaces. Water absorbs nearly all solar radiation in the SWIR (1000–2500 nm), resulting in minimal reflectance, and may alter plastic signatures unless the material is floating or dry (Garaba & Dierssen, 2020; Moshtaghi et al., 2021).

Outside of the agricultural sector, we also identify HDPE in a recyclable site in Almería, Spain (37°24′41.9″N 1°48′54.1″W) as shown in Figure 4(a). Our findings underscore EMIT's ability to detect waste management practices and potential sources of plastic pollution in terrestrial environments, which represent two key components of the plastic cycle. These components among other numerous processes of the cycle undergo a series of steps ultimately leading to pathways toward waste containment, storage, or greenhouse gas emissions (Zhu, 2021). They represent only the initial stages of tracking a more complex cycle introduced by the Anthropocene era, which has become an integral part of the carbon cycle as plastic originates from fossil carbon. In addition, we have also detected plastic material suggestive of HDPE in the roof of a warehouse of an aquafarm, as shown in Figure 4(b).

While EMIT's broad coverage begins to uncover the broad extent of surface plastics, its non-global coverage and ground sampling that is coarse relative to many plastic generating activities mean that more work is needed to understand the global distribution of surface plastics. However, this early study provides important new information about the appearance of hydrocarbon signatures across large parts of Earth's terrestrial surface area. To our knowledge, it represents the first direct detection of terrestrial plastics on continental scales. Additional considerations are needed to use this information effectively in enhancing our comprehension of global plastic sources. Limitations remain in the detection of macroplastics, as some common plastic polymers lack spectrally distinct absorption features, even when using imaging spectroscopy. Furthermore, many instances of plastic pollution may not be reliably detected, either because the material is buried below the surface, as is often the case with landfills, or because the spatial extent is too small to be resolved in 60-m data, even with a matched filter approach. Also, some of the plastic detections may be transient, meaning that they are not in the environment for long periods or were not present during the time of the EMIT overpass. Therefore, further analysis using observations from EMIT's extended mission will help to evaluate temporal variations in plastic detections and how these relate to things such as seasonal agricultural activities. Extended observations and broader coverage will also be crucial for validating the initial trends observed in our global plastic mapping. This is especially important given that the data coverage in this work does not include China, the world's largest producer and consumer of plastics (Luan et al., 2021). Validation of EMIT observed trends would also greatly benefit from the synergy with other complementary satellite missions, including those from international partners like the European Space Agency's Copernicus Sentinels. Additionally, not all detection of surface plastics from imaging spectroscopy observations may be equally indicative of substrates that are likely to contribute to plastic debris entering the environment. Such signatures are unlikely to represent a major source of plastic pollution, and could be distinguished from other plastic sources using their specific spectral features.

4 Conclusion

To improve the identification of land-based sources that can lead to potential contamination of the soil and oceans, we have used the EMIT orbital imaging spectrometer to detect plastic signatures at continental scales. We have demonstrated EMIT's ability to detect plastics, identifying High-Density Polyethylene (HDPE) and Polyvinyl Chloride (PVC) across vast areas of Earth's terrestrial surface within just a 4-month observation period. Notably, we have detected hot spots of these plastics in regions such as Europe, the United States, and East Asian countries like Taiwan, South Korea, Japan, and the Philippines. These detections show the extensive use of plastics in agricultural practices, including hoop houses and aquaculture. While our results are constrained by the limited coverage of the chosen time period, further analysis incorporating observations from EMIT's extended mission will offer broader coverage to validate the initial detection trends identified in this study. Furthermore, future spectrometers, such as the Surface Biology and Geology (SBG) and Copernicus Hyperspectral Imaging Mission for the Environment (CHIME) missions, will provide regular pole-to-pole coverage, allowing a more comprehensive survey of surface plastic presence and dynamics.