Open-access Satellite Data Research Articles

Forest age estimation using UAV-LiDAR and Sentinel-2 data with machine learning algorithms- a case study of Masson pine (Pinus massoniana)

ABSTRACT The assessment of ecological functions, such as those of forest structure zoning and carbon sinks, heavily relies on forest age classification. Therefore, monitoring forest age is a crucial element of forest resource surveys. With the increased availability of high-quality open-access satellite data and advancements in Unmanned Aerial Vehicle Light Detection and Ranging (UAV-LiDAR) technology, remote sensing has emerged as an essential method for acquiring accurate forest age information. In this study, Sentinel-2 remote sensing data, UAV-LiDAR data, and combined Sentinel-2 and LiDAR data are used as data sources. Three machine learning algorithms, Adaptive Boosting (AdaBoost), Random Forest (RF), and Extreme Random Tree (ERT), are used to predict forest age in a Masson pine (Pinus massoniana Lamb.) forest. The optimal model is used to predict the forest age and simulate the spatial age distribution. The machine learning models based on separate Sentinel-2 and LiDAR data accurately predict the age of the Masson pine forest. Nevertheless, the accuracy of the RF model with combined data was higher than that in other cases, with an accuracy R value of 0.81. The model displayed good stability, and the spatial uncertainty of age estimation was low. Compared with the RF model using only Sentinel-2 data (R = 0.43), the RF model with combined LiDAR and Sentinel-2 data achieved the highest accuracy, with R values 88.37% higher. In addition, the forest canopy structure parameters, such as the average height of the forest stand extracted from UAV-LiDAR data, had a significant impact on the estimation of forest age. Thus, when the combined Sentinel-2 and LiDAR data were used to establish these parameters, the highest accuracy in the estimation of Masson pine was obtained. The findings of this study provide new insights for forest age estimation based on multi-source remote sensing data.

European vineyards and their cultural landscapes exposed to record drought and heat

CONTEXTEuropean vineyards, producing over 50% of the world's wine and hosting significant cultural landscapes, face threats from climate change and related severe weather events. The summer of 2022, especially July, posed significant challenges due to extreme drought and high temperatures. However, this period also provided an opportunity to study how such events might involve viticulture in Europe and to explore mitigation solutions. OBJECTIVEThe objectives are (1) to characterize the severity of the extreme event of July 2022 in European wine regions regarding primary climatic parameters, (2) map vineyards at risk due to agricultural drought and high land surface temperature, and (3) discuss the role of various water-related interventions for mitigating similar events. METHODSAfter identifying the locations of European vineyards using the Corine Land Cover 2018 (CLC2018), open-access satellite data were employed to: (1) assess anomalies in Maximum Air Temperature (NATm), Land Surface Temperature (LST), Precipitation (P), and Soil Moisture (SM) in July 2022 compared to long-term averages; (2) identify regions at higher risk that experienced extreme agricultural drought (Vegetation Health Index, VHI = Extreme) and LST > 35 °C. RESULTS AND CONCLUSIONSIn July 2022, European vineyards experienced an average increase of 11% in NATm, a 9% rise in LST, and a reduction of 47% in P and 30% in SM compared to historical averages. 18% of European vineyards were at risk of drought and excessive heat, particularly in Portugal (31%), France (27%), and Italy (21%), including 10 viticultural cultural landscapes. Findings highlight the urgent need for long-term sustainable water management practices over emergency interventions. This research supports informed decision-making, emphasizing that climate resilience is necessary for preserving the cultural heritage of European wine-growing areas. SIGNIFICANCEThis research provides an overview of the dynamics of extreme events on viticulture at a continental scale, promoting climate-aware viticultural systems and offering insight scalable to global viticulture under changing climatic conditions.

Seagrasses on the move: Tracing the multi-decadal species distribution trends in lagoon meadows using Landsat imagery

Seagrass meadows play a vital role for lagoon ecosystems and their biota, sustaining multiple ecosystem services. Their distribution and functioning are closely tied to the environmental pressures induced by global changes. Long-term monitoring of seagrass species and communities is, hence, important to depict their response to past and future scenarios. The availability of long term open-access satellite data offers a new remote sensing perspective for monitoring seagrass communities dynamics in shallow waters, especially when combined with machine learning algorithms. In this study, seasonal multispectral images (from 1999 to 2019) were collected from Landsat 5 Thematic Mapper and 8 Operational Land Imager satellites to map the seagrass meadows, at the community and species levels, within the vast Grado and Marano lagoon (Northeast Italy) using a Random Forest (RF) algorithm. RF models were calculated using an extensive field training dataset collected in 2010 (n = 426) and reached an overall accuracy of 0.92 and 0.76 for the classification at the community and species levels, respectively. The change detection analysis revealed an increase of 14.16 km2 (+ 39%) of the whole seagrass community cover over the period, at a rate of 1.59 km2year−1. Despite the coarse spatial resolution (30 m) of the Landsat's images, the classification of seagrasses at species level achieved a good overall accuracy (0.76), evidencing Nanozostera noltei as the species with the highest cover increase (+13.87 km2 over the time period). The observed expansion is likely caused by an increase of the sea water influence that is radically modifying Adriatic brackish water bodies, emphasizing the connection between the ongoing environmental changes and the rapid responses of seagrass meadows.

Observations of Archaeological Proxies through Phenological Analysis over the Megafort of Csanádpalota-Juhász T. tanya in Hungary Using Sentinel-2 Images

This study aims to investigate potential archaeological proxies at a large Bronze Age fortification in Hungary, namely the Csanádpalota–Juhász T. tanya site, using open-access satellite data. Available Sentinel-2 images acquired between April 2017 and September 2022 were used. More than 700 images (727) were initially processed and filtered, accounting at the end of more than 400 (412) available calibrated Level 2A Sentinel images over the case study area. Sentinel-2 images were processed through image analysis. Based on pan-sharpened data, the visibility of crop marks was improved and enhanced by implementing orthogonal equations. Several crop marks, some still unknown, were revealed in this study. In addition, multi-temporal phenological observations were recorded on three archaeological proxies (crop marks) within the case study area, while an additional area was selected for calibration purposes (agricultural field). Phenological observations were performed for at least four complete phenological cycles throughout the study period. Statistical comparisons between the selected archaeological proxies were applied using a range of vegetation indices. The overall results indicated that phenological observations could be used as archaeological proxies for detecting the formation of crop marks.

Predicting Species and Structural Diversity of Temperate Forests with Satellite Remote Sensing and Deep Learning

Anthropogenically-driven climate change, land-use changes, and related biodiversity losses are threatening the capability of forests to provide a variety of valuable ecosystem services. The magnitude and diversity of these services are governed by tree species richness and structural complexity as essential regulators of forest biodiversity. Sound conservation and sustainable management strategies rely on information from biodiversity indicators that is conventionally derived by field-based, periodical inventory campaigns. However, these data are usually site-specific and not spatially explicit, hampering their use for large-scale monitoring applications. Therefore, the main objective of our study was to build a robust method for spatially explicit modeling of biodiversity variables across temperate forest types using open-access satellite data and deep learning models. Field data were obtained from the Biodiversity Exploratories, a research infrastructure platform that supports ecological research in Germany. A total of 150 forest plots were sampled between 2014 and 2018, covering a broad range of environmental and forest management gradients across Germany. From field data, we derived key indicators of tree species diversity (Shannon Wiener Index) and structural heterogeneity (standard deviation of tree diameter) as proxies of forest biodiversity. Deep neural networks were used to predict the selected biodiversity variables based on Sentinel-1 and Sentinel-2 images from 2017. Predictions of tree diameter variation achieved good accuracy (r2 = 0.51) using Sentinel-1 winter-based backscatter data. The best models of species diversity used a set of Sentinel-1 and Sentinel-2 features but achieved lower accuracies (r2 = 0.25). Our results demonstrate the potential of deep learning and satellite remote sensing to predict forest parameters across a broad range of environmental and management gradients at the landscape scale, in contrast to most studies that focus on very homogeneous settings. These highly generalizable and spatially continuous models can be used for monitoring ecosystem status and functions, contributing to sustainable management practices, and answering complex ecological questions.

Nearshore Bathymetry Based on ICESat-2 and Multispectral Images: Comparison Between Sentinel-2, Landsat-8, and Testing Gaofen-2

Accurate bathymetric maps are essential to understand marine and coastal ecosystems. With the development of satellite and sensor technology, satellite-derived bathymetry (SDB) has been widely used to measure the depth of nearshore waters. Employment of physics-based methods requires a series of optical parameters of the water column and seafloor, which limits the application of these methods to shallow-water bathymetry. Due to convenience, low costs, and high efficiency, empirical methods based on <i>in situ</i> measurements and satellite imagery are increasingly used for nearshore bathymetry. These measurements are required to calibrate empirical models, so that reasonable accuracy can be achieved. The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2), equipped with the advanced topographic laser altimeter system, provides a novel opportunity for nearshore bathymetry. Using the new measurement strategy of photon counting, ICESat-2 can provide accurate bathymetric points from spaceborne observations, which can be used in place of <i>in situ</i> water depth data. In this study, ICESat-2 bathymetric points and multispectral images were used to train four typical models and produce bathymetric maps for Shanhu Island, Ganquan Island, and Lingyang Reef in the Xisha Islands of China. We evaluated the bathymetric results by comparing them with reference depth data from airborne light detection and ranging. All models had a satisfactory accuracy, as well as multimodel and multisource image consistency. With the ICESat-2 bathymetric points, SDB is no longer limited by <i>in situ</i> measurements. Hence, this approach could be extended to a larger scale to obtain nearshore bathymetric maps of coastal areas, surrounding islands, and reefs using free and open-access satellite data.

Оценка сезонного водного баланса бассейна реки Абай в Эфиопии с использованием спутниковых баз данных и специализированной гидрологической модели

The paper examines the seasonal water balance of the Abay River basin (Ethiopia) using open-access satellite data and a hydrological model representing the formation of water resources. The conducted analysis and theoretical studies indicate that hydrological modeling and remote sensing are the most appropriate tools for assessing water balance, which is essential for determining water prices and op-timizing water resource management. In this work, QGIS models were applied to estimate water deficit during the dry season and water surplus during the rainy season, using those data to determine sea-sonal water balance for six years. To this end, the study employed GeoTIFF metadata and additional information, including a map projection, coordinate systems, ellipsoids, and baseline data necessary to establish the accurate spatial reference of the file. The water balance analysis required data on evapo-ration, precipitation, surface runoff, and surface water reserves in the Abay River basin. According to the calculation results, a significant surplus of water was noted during the rainy season and its deficit during the dry season. The obtained numerical data were used to model the timing of droughts and floods, as well as to manage urban drainage systems and agricultural irrigation.

Application of Machine Learning on Google Earth Engine to Produce Landslide Susceptibility Mapping (Case Study: Pacitan)

According to the Indonesian Disaster Management Agency (BNPB), Indonesia’s losses due to landslides were estimated around hundreds of billion rupiah in 2017. Making landslide as one of the catastrophes with the greatest risk of loss and leaving a couple regions prone to landslides in Indonesia, Pacitan region is one of them. Landslide delineation therefore represents a particularly beneficial application of evolving research trend in disaster reduction, especially for the vulnerable region. In the present times of open-access satellite data, cloud computing and machine-learning algorithms is frequently used for disaster prevention monitoring. By employing Google Earth Engine, this study focuses on the susceptibility of landslide occurrence using a random forest machine-learning framework applied to digital topographic data such as elevation, slope and aspect as the independent variables and landslide inventory data obtained from Ministry of Energy and Mineral Resources Republic of Indonesia as the dependent variable. This study data sets composed from 1000 random points in Pacitan region with 70:30 ratio for training and testing sample points. The model produced good result, with overall accuracy values of 0.94, kappa values of 0.79 and 0.80 for AUC value. This model also showed that elevation is the most important variable in the landslide susceptible area. The results of this study can be used to evaluate the potential future impacts of landslide and help to optimize the management of disaster reduction in the region of Pacitan.

Mapping Coastal Wetlands of the Bohai Rim at a Spatial Resolution of 10 m Using Multiple Open-Access Satellite Data and Terrain Indices

Coastal wetlands provide essential ecosystem services and are closely related to human welfare. However, they can experience substantial degradation, especially in regions in which there is intense human activity. To control these increasingly severe problems and to develop corresponding management policies in coastal wetlands, it is critical to accurately map coastal wetlands. Although remote sensing is the most efficient way to monitor coastal wetlands at a regional scale, it traditionally involves a large amount of work, high cost, and low spatial resolution when mapping coastal wetlands at a large scale. In this study, we developed a workflow for rapidly mapping coastal wetlands at a 10 m spatial resolution, based on the recently emergent Google Earth Engine platform, using a machine learning algorithm, open-access Synthetic Aperture Radar (SAR) and optical images from the Sentinel satellites, and two terrain indices. We then generated a coastal wetland map of the Bohai Rim (BRCW10) based on the workflow. It has a producer accuracy of 82.7%, according to validation using 150 wetland samples. The BRCW10 data reflected finer information when compared to wetland maps derived from two sets of global high-spatial-resolution land cover data, due to the fusion of multiple data sources. The study highlights the benefits of simultaneously merging SAR and optical remote sensing images when mapping coastal wetlands.

Water, water, but not everywhere: analysis of shrinking water bodies using open access satellite data

ABSTRACT The importance of water cannot be overstated. Not only it is essential for life, it is a vital resource for sustainable development. Water resource management is pivotal to ensure water security and sustainability. Having such water monitoring resources is challenging, especially in less developed countries that have limited resources. With the advancement in technology and availability of open-access satellite data, there are opportunities to reinforce the abilities of researchers and decision-makers to monitor and identify changes in water resources. In this work, we describe an easy to implement workflow for estimation of surface area of waterbodies using open-access data from Sentinel-2 satellites. The method uses Normalized Difference Water Index (NDWI) for surface water detection that is used to estimate the surface area of the waterbody and analyse changes over a period of time. We develop four case studies based on waterbodies in Spain, Morocco, and India to highlight how satellite imaging can effectively help in estimating and visualizing the changes in water surface area. Temperature and rainfall data are analysed to understand the impact of climate change on water resources due to increase in temperature and changes in precipitation patterns. The proposed method is validated by testing it on four additional lakes and comparing the results with publicly available data. We show how open-access satellite data can potentially be used for regular monitoring of waterbodies as well as filling the knowledge gap to assist in decision making at both local and regional levels.

Satellite-derived bathymetry using the ICESat-2 lidar and Sentinel-2 imagery datasets

Accurate bathymetric data is essential for marine, coastal ecosystems, and related studies. In the past decades, a lot of studies were investigated to obtain bathymetric data in shallow waters using satellite remotely sensed data. Satellite multispectral imagery has been widely used to estimate shallow water depths based on empirical models and physics-based models. However, the in-situ water depth information is essential (as the priori) to use the empirical model in a specific area, which limits its application, especially for remote reefs. In this study, the bathymetric maps in shallow waters were produced based on empirical models with only satellite remotely sensed data (i.e., the new ICESat-2 bathymetric points and Sentinel-2 multispectral imagery). The bathymetric points from the spaceborne ICESat-2 lidar were used in place of the in-situ auxiliary bathymetric points to train the classical empirical models (i.e., the linear model and the band ratio model). The bathymetric points were firstly extracted from noisy ICESat-2 raw data photons by an improved point cloud processing algorithm, and then were corrected for bathymetric errors (which were caused by the refraction effect in the water column, the refraction effect on the water surface, and the fluctuation effect on the water surface). With the trained empirical models and Sentinel-2 multispectral images, the bathymetric maps were produced for Yongle Atoll, in the South China Sea and the lagoon near Acklins Island and Long Cay, to the southeast of Bahama with four-date Sentinel-2 images. The bathymetry performance (including the accuracy and consistency of multi-date data) was evaluated and compared with the in-situ measurements. The results indicate that the bathymetric accuracy is well, and the RMSE is lower or close to 10% of the maximum depth for the two models with four-date images in two study areas. The consistency of multi-date data is well with the mean R2 of 0.97. The main novelties of this study are that the accuracy bathymetric points can be obtained from the ICESat-2 raw data using the proposed signal processing and error correction method, and using the ICESat-2 bathymetric points, the satellite multispectral imagery based on empirical models is no longer limited by local priori measurements, which were essential in previous studies. Hence, In the future, with the help of free and open-access satellite data (i.e., ICESat-2 data and Sentinel-2 imagery), this approach can be extended to a larger scale to obtain bathymetric maps in the shallow water of coastal areas, surroundings of islands and reefs, and inland waters.

Green Apartheid: Urban green infrastructure remains unequally distributed across income and race geographies in South Africa

Urban green infrastructure provides ecosystem services that are essential to human wellbeing. A dearth of national-scale assessments in the Global South has precluded the ability to explore how political regimes, such as the forced racial segregation in South Africa during and after Apartheid, have influenced the extent of and access to green infrastructure over time. We investigate whether there are disparities in green infrastructure distributions across race and income geographies in urban South Africa. Using open-source satellite imagery and geographic information, along with national census statistics, we find that public and private green infrastructure is more abundant, accessible, greener and more treed in high-income relative to low-income areas, and in areas where previously advantaged racial groups (i.e. White citizens) reside. Areas with White residents report 6-fold higher income, have 11.7% greater tree cover, 8.9% higher vegetation greenness and live 700 m closer to a public park than areas with predominantly Black African, Indian, and Coloured residents. The inequity in neighborhood greenness levels has been maintained (for Indian and Coloured areas) and further entrenched (for Black African areas) since the end of Apartheid in 1994 across the country. We also find that these spatial inequities are mirrored in both private (gardens) and public (street verges, parks, green belts) spaces, hinting at the failure of governance structures to plan for and implement urban greening initiatives. By leveraging open-access satellite data and methods presented here, there is scope for civil society to monitor urban green infrastructure over time and thereby hold governments accountable to addressing environmental justice imperatives in the future. Interact with the data here: green-apartheid.zsv.co.za.

Generating intra-year metrics of wildfire progression using multiple open-access satellite data streams

The 2017 fire season was one of the largest on record for British Columbia (BC), Canada, in terms of total area burned (estimated 1.2 million ha), affecting the safety and air quality of numerous communities. Moreover, fires of this number and extent alter the wood supply for harvesting, the nature of habitat for wildlife, and can affect regional and national carbon budgets. As a result, it is important to map these fires accurately and to monitor within-year fire progression in order to quantify the resulting forest-disturbance impacts fully. The Bayesian Updating of Land Cover (BULC) algorithm was used to merge burned-area classifications of individual fires from a range of remote sensing sources such as Landsat-7, Landsat-8, Sentinel-2, and MODIS (MCD64A1) burned-area dataset. Together, these provisional classifications imaged each pixel within a known fire perimeter an average of 33.8 times between April 1 and December 1. The resulting 35-week time-series stack had updated weekly burned areas for each of the 89 fires in BC in 2017. Province-wide fire progression was variable throughout the period analyzed, characterized by a steady burn phase of 41,437 ha (5% of total area burned) over a two-week period in early July, an accelerated burn phase of 149,422 ha (17%) from mid-July to early August, another steady burn phase of 218,079 ha (24%) for one month until early September, and a second accelerated burn phase of 301,931 ha (34%) over two weeks in late September with subsequent steady growth of 180,119 ha (20%) over 1.5 months until containment in late October. Herein, we demonstrate how such temporally dense fire classification stacks can be used to analyze fire progression over the course of a fire season (both retrospectively and in near-real time) providing useful metrics to characterize and compare fire events. End-of-season burned-area estimates correspond with estimates derived from the National Burned Area Composite (NBAC) product that is generated retrospectively from fire best-available mapping approaches. This rapid interpretation of information enables the analysis of suppression success and potential drivers of fires spread while facilitating analyses of carbon budget consequences as well as impacts to communities and timber supply.

Large-scale probabilistic identification of boreal peatlands using Google Earth Engine, open-access satellite data, and machine learning.

Freely-available satellite data streams and the ability to process these data on cloud-computing platforms such as Google Earth Engine have made frequent, large-scale landcover mapping at high resolution a real possibility. In this paper we apply these technologies, along with machine learning, to the mapping of peatlands–a landcover class that is critical for preserving biodiversity, helping to address climate change impacts, and providing ecosystem services, e.g., carbon storage–in the Boreal Forest Natural Region of Alberta, Canada. We outline a data-driven, scientific framework that: compiles large amounts of Earth observation data sets (radar, optical, and LiDAR); examines the extracted variables for suitability in peatland modelling; optimizes model parameterization; and finally, predicts peatland occurrence across a large boreal area (397, 958 km2) of Alberta at 10 m spatial resolution (equalling 3.9 billion pixels across Alberta). The resulting peatland occurrence model shows an accuracy of 87% and a kappa statistic of 0.57 when compared to our validation data set. Differentiating peatlands from mineral wetlands achieved an accuracy of 69% and kappa statistic of 0.37. This data-driven approach is applicable at large geopolitical scales (e.g., provincial, national) for wetland and landcover inventories that support long-term, responsible resource management.

Open-access cloud resources contribute to mainstream REDD+: The case of Mozambique

The objective of this work is to investigate how Earth Observation data and processing platforms accessible on the cloud can facilitate the implementation of REDD+ in developing countries. For that, we explore newly available open-access satellite data, cloud processing, and a ready-made land cover map to assess the extent to which such resources can directly respond to monitoring and measuring, reporting, and verification requisites. Mozambique is one of the 47 countries selected to benefit from the Forest Carbon Partnership Facility to implement REDD+ strategies. However, to meet funding agreements, the country needs to periodically produce national land cover and land cover change maps at given resolution and accuracy levels. The work presented here shows that the land cover mapping requisites of REDD+ may be quickly and cost-effectively met through the development and use of newly available cloud resources. The study relies on an experimental design that tests the results of image processing approaches with algorithms available or developed in Google Earth Engine against country-wide reference data collected by a team of national experts. The results show that, in addition to pre-processing advantages, which facilitate multi-temporal compositing and mosaicking of very large and heavy data sets, developments in cloud processing and image classification swiftly produce large extent and high-resolution land cover maps, tailored to a specific objective. The comparison of results between the in-house map obtained using Google Earth Engine, and the pan-African map produced by the European Space Agency (2016) at the same spatial resolution, reveals that both maps meet REDD+ requirements for a binary Forest/Non-forest legend. However, the in-house map is more accurate and reaches considerably better results if a more complex, six class IPCC legend is required. Nevertheless, this study shows that, given adequate reference data, the need to periodically produce high resolution land cover maps for national forest monitoring purposes is no longer an obstacle for mainstreaming the implementation of REDD+.

Google Earth Engine, Open-Access Satellite Data, and Machine Learning in Support of Large-Area Probabilistic Wetland Mapping

Modern advances in cloud computing and machine-leaning algorithms are shifting the manner in which Earth-observation (EO) data are used for environmental monitoring, particularly as we settle into the era of free, open-access satellite data streams. Wetland delineation represents a particularly worthy application of this emerging research trend, since wetlands are an ecologically important yet chronically under-represented component of contemporary mapping and monitoring programs, particularly at the regional and national levels. Exploiting Google Earth Engine and R Statistical software, we developed a workflow for predicting the probability of wetland occurrence using a boosted regression tree machine-learning framework applied to digital topographic and EO data. Working in a 13,700 km2 study area in northern Alberta, our best models produced excellent results, with AUC (area under the receiver-operator characteristic curve) values of 0.898 and explained-deviance values of 0.708. Our results demonstrate the central role of high-quality topographic variables for modeling wetland distribution at regional scales. Including optical and/or radar variables into the workflow substantially improved model performance, though optical data performed slightly better. Converting our wetland probability-of-occurrence model into a binary Wet-Dry classification yielded an overall accuracy of 85%, which is virtually identical to that derived from the Alberta Merged Wetland Inventory (AMWI): the contemporary inventory used by the Government of Alberta. However, our workflow contains several key advantages over that used to produce the AMWI, and provides a scalable foundation for province-wide monitoring initiatives.

Using Open Access Satellite Data Alongside Ground Based Remote Sensing: An Assessment, with Case Studies from Egypt’s Delta

This paper will assess the most recently available open access high-resolution optical satellite data (0.3 m–0.6 m) and its detection of buried ancient features versus ground based remote sensing tools. It also discusses the importance of CORONA satellite data to evaluate landscape changes over the past 50 years surrounding sites. The study concentrates on Egypt’s Nile Delta, which is threatened by rising sea and water tables and urbanization. Many ancient coastal sites will be lost in the next few decades, thus this paper emphasizes the need to map them before they disappear. It shows that high resolution satellites can sometimes provide the same general picture on ancient sites in the Egyptian Nile Delta as ground based remote sensing, with relatively sandier sedimentary and degrading tell environments, during periods of rainfall, and higher groundwater conditions. Research results also suggest potential solutions for rapid mapping of threatened Delta sites, and urge a collaborative global effort to maps them before they disappear.