scholarly journals Highly Local Model Calibration with a New GEDI LiDAR Asset on Google Earth Engine Reduces Landsat Forest Height Signal Saturation

2020 ◽  
Vol 12 (17) ◽  
pp. 2840 ◽  
Author(s):  
Sean P. Healey ◽  
Zhiqiang Yang ◽  
Noel Gorelick ◽  
Simon Ilyushchenko

While Landsat has proved to be effective for monitoring many elements of forest condition and change, the platform has well-documented limitations in measuring forest structure, the vertical distribution of the canopy. This is important because structure determines several key ecosystem functions, including: carbon storage; habitat suitability; and timber volume. Canopy structure is directly measured by LiDAR, and it should be possible to train Landsat structure models at a highly local scale with the dense, global sample of full waveform LiDAR observations collected by NASA’s Global Ecosystem Dynamics Investigation (GEDI). Local models are expected to perform better because: (a) such models may take advantage of localized correlations between structure and canopy surface reflectance; and (b) to the extent that models revert to the mean of the calibration data due to a lack of discrimination, local models will revert to a more representative mean. We tested Landsat-based relative height predictions using a new GEDI asset on Google Earth Engine, described here. Mean prediction error declined by 23% and important prediction biases at the extremes of the range of canopy height dropped as model calibration became more local, minimizing forest structure signal saturation commonly associated with Landsat and other passive optical sensors. Our results suggest that Landsat-based maps of structural variables such as height and biomass may substantially benefit from the kind of local calibration that GEDI’s dense sample of LiDAR data supports.

2021 ◽  
Author(s):  
Wahaj Habib ◽  
John Connolly ◽  
Kevin McGuiness

<p>Peatlands are one of the most space-efficient terrestrial carbon stores. They cover approximately 3 % of the terrestrial land surface and account for about one-third of the total soil organic carbon stock. Peatlands have been under severe strain for centuries all over the world due to management related activities. In Ireland, peatlands span over approximately 14600 km<sup>2</sup>, and 85 % of that has already been degraded to some extent. To achieve temperature goals agreed in the Paris agreement and fulfil the EU’s commitment to quantifying the Carbon/Green House Gases (C/GHG) emissions from land use, land use change forestry, accurate mapping and identification of management related activities (land use) on peatlands is important.</p><p>High-resolution multispectral satellite imagery by European Space Agency (ESA) i.e., Sentinel-2 provides a good prospect for mapping peatland land use in Ireland. However, due to persistent cloud cover over Ireland, and the inability of optical sensors to penetrate the clouds makes the acquisition of clear sky imagery a challenge and hence hampers the analysis of the landscape. Google Earth Engine (a cloud-based planetary-scale satellite image platform) was used to create a cloud-free image mosaic from sentinel-2 data was created for raised bogs in Ireland (images collected for the time period between 2017-2020). A preliminary analysis was conducted to identify peatland land use classes, i.e., grassland/pasture, crop/tillage, built-up, cutover, cutaway and coniferous, broadleaf forests using this mosaicked image. The land-use classification results may be used as a baseline dataset since currently, no high-resolution peatland land use dataset exists for Ireland. It can also be used for quantification of land-use change on peatlands. Moreover, since Ireland will now be voluntarily accounting the GHG emissions from managed wetlands (including bogs), this data could also be useful for such type of assessment.</p>


Author(s):  
Muhammad Ali Ismail ◽  
Maria Waqas ◽  
Amjad Ali ◽  
Mirza Muhammad Muzzamil ◽  
Uzair Abid ◽  
...  

Abstract The sustainability of the hydrological and ecological ecosystems of any region requires continuous monitoring of the water bodies. Recent advancements in satellite-based remote optical sensors, big data analysis and cloud computing have given new dimensions to the field of water body studies including their detection as well as analysis. The present study extends the existing methods to assess the contemporary surface water detection and monitoring techniques via remote sensing. The proposed technique implies an improved hybrid approach for the purpose along with the calculation of the boundary areas. The study has been carried out on the Manchar Lake, the largest natural freshwater lake in Pakistan as well as in South Asia. The proposed hybrid water index along with the different existing water body detection indices and spectral bands have been worked out on the satellite images retrieved from the Google Earth Engine to detect and analyze the area/flow changes in the water body. Based on the 7 years of data, the proposed algorithm calculates the water body area more precisely. With limited availability of metadata about the study area, the results have been validated both qualitatively through national-met data and statistically. These results aid to better preserve and improve the quality of the water resource.


2021 ◽  
Author(s):  
Renata Rossoni ◽  
Fernando Fan ◽  
Leonardo Laipelt

<p>In hydrosedimentological modelling, the lack of high temporal resolution field data is a limiting factor for the assessment of the performance of models. This way, the remote sensing images have been studied to correlate imagery information with suspended sediment concentration (SSC) in the last decades, aiming to complement field data, by improving the SSC information temporal and spatial resolution. Thus, the present work used the Google Earth Engine (GEE), a cloud-based platform, to systematically obtain red band reflectance information from Landsat 5 imagery to support large-scale hydrosedimentological modelling. The test case was to the Rio Grande do Sul state hydrological region in Brazil, a South American region with scarce SSC data. The methodology applied consisted in nine steps using GEE code: (1) river width analysis using remote sensing imagery to localize the virtual gauge stations (VGS) from the intersection between the discretization of hydrosedimentological model and the chosen rivers, (2) TM sensor definition, onboard of Landsat 5 satellite, (3) collection of red reflectance information between 1990 and 2010, based on previous works that presented better correlation between red reflectance and SSC, (4) in each VGS, we created a circle of radius equal to 1000 m, (5) to each image, we removed clouded-pixels, using the Landsat 5 quality bands, (6) we generated a dynamic water mask to each image to ensure that only pixels with water would be used to collected reflectance information, (7) finally, we calculated the mean of red band reflectance inside the intersection of water mask and circle buffer, removing the clouded-pixels, (8) we calculated a filter to remove remnants clouded-pixels and random errors from imagery, (9) we used the MGB-SED model to simulate long-term SSC in the region and we calibrated the model with the GEE data based on a correlation approach. The results found were: (i) 1267 virtual gauge stations, approximately 20 times the number of in situ SSC gauging stations available in the region, (ii) a larger area of data and greater temporal resolution, (iii) improvement in the correlation between model results and red reflectance, when we assess the model with SSC observed data. In conclusion, the work shows the potential of GEE to simply obtain large-scale reflectance data that could be used to improve the calibration processes of large-scale hydrosedimentological modelling.</p>


Author(s):  
Mauricio Vega-Araya

La Tierra y su biosfera están cambiando constantemente, por lo tanto, es fundamental detectar los cambios con el fin de entender su impacto en los ecosistemas terrestres. Los esquemas de monitoreo de ecosistemas han evolucionado rápidamente en las ultimas décadas. En el caso del monitoreo forestal, los métodos y herramientas que facilitan la utilización de imágenes satelitales permiten realizar este monitoreo con el cual se puede detectar donde y cuando un bosque es eliminado o afectado debido a un evento de deforestación o bien de fuego, lo anterior casi en tiempo real. Estas nuevas herramientas están disponibles para su implementación, sin embargo, ningún paı́s de la región centroamericana y el Caribe ha implementado un sistema como herramienta de decisión dentro de una estructura de gobierno central o federal debido a la ausencia de programas de transferencia de tecnologı́a o programas de capacitación de talento local. Los sensores remotos proporcionan mediciones consistentes y repetibles que permiten la captura de los efectos de muchos procesos que causan el cambio, incluyendo, por ejemplo, incendios, ataques de insectos, agentes de cambio naturales y antropogénicas como por ejemplo, la deforestación, la urbanización, la agricultura, etc. Las series temporales de imágenes de satélite proporcionan maneras para detectar y vigilar cambios en el tiempo y en el espacio, esto consistentemente durante los últimos 30 años a nivel mundial. Los incendios forestales afectan el proceso de sucesión del bosque, no obstante, es muy limitada la existencia de estudios locales que relacionen el efecto de los incendios forestales con las diferencias en la información espectral a partir de sensoramiento remoto. En el presente estudio se plantea y propone la utilización y aprovechamiento de lo que se ha denominado grandes datos, especialmente con el advenimiento muchas plataformas de sensores remotos como Landsat, MODIS y recientemente Sentinel, para identificar cuál es el efecto de los incendios forestales en la sucesión y sus elementos perturbadores, como por ejemplo, la presencia de lianas. Se procesaron las series temporales se usó la plataforma digital Google Earth Engine, que permitió la selección y reducción de la información espacial de los ı́ndices de vegetación en tendencia, estacionalidad y residuos. Se analizó la respuesta de estos ı́ndices en sitios con diferente afectación por incendios forestales. Con estos índices se pretende desarrollar modelos de clasificación de series espaciales de tiempo de los ı́ndices y poder ası́ comprender los cambios en el tiempo y el espacio de los ecosistemas afectados por incendios forestales. Preliminarmente, se encontró una relación entre la incidencia de los incendios forestales y el fenómeno del Niño-Oscilación del Sur para el índice de vegetación denominado índice de área foliar. Además, la evidencia indica que el índice normalizado de vegetación si presenta diferencias respecto a los sitios que tienen un historial de fuegos diferente. El establecer esta relación implica estudiar también los regı́menes de precipitación y temperatura. El descomponer las series de tiempo facilitó la correlación con otras series de tiempo, permitiendo establecer las bases de un monitoreo y a su vez, relacionar las índices de vegetación y su variación con otros elementos climáticos, como por ejemplo, el efecto ENOS.


2018 ◽  
Vol 54(9) ◽  
pp. 29
Author(s):  
Võ Quốc Tuấn ◽  
Nguyễn Thiên Hoa ◽  
Huỳnh Thị Kim Nhân ◽  
Đặng Hoàng Khải

2021 ◽  
Vol 13 (8) ◽  
pp. 1433
Author(s):  
Shobitha Shetty ◽  
Prasun Kumar Gupta ◽  
Mariana Belgiu ◽  
S. K. Srivastav

Machine learning classifiers are being increasingly used nowadays for Land Use and Land Cover (LULC) mapping from remote sensing images. However, arriving at the right choice of classifier requires understanding the main factors influencing their performance. The present study investigated firstly the effect of training sampling design on the classification results obtained by Random Forest (RF) classifier and, secondly, it compared its performance with other machine learning classifiers for LULC mapping using multi-temporal satellite remote sensing data and the Google Earth Engine (GEE) platform. We evaluated the impact of three sampling methods, namely Stratified Equal Random Sampling (SRS(Eq)), Stratified Proportional Random Sampling (SRS(Prop)), and Stratified Systematic Sampling (SSS) upon the classification results obtained by the RF trained LULC model. Our results showed that the SRS(Prop) method favors major classes while achieving good overall accuracy. The SRS(Eq) method provides good class-level accuracies, even for minority classes, whereas the SSS method performs well for areas with large intra-class variability. Toward evaluating the performance of machine learning classifiers, RF outperformed Classification and Regression Trees (CART), Support Vector Machine (SVM), and Relevance Vector Machine (RVM) with a >95% confidence level. The performance of CART and SVM classifiers were found to be similar. RVM achieved good classification results with a limited number of training samples.


2021 ◽  
Vol 13 (12) ◽  
pp. 2299
Author(s):  
Andrea Tassi ◽  
Daniela Gigante ◽  
Giuseppe Modica ◽  
Luciano Di Martino ◽  
Marco Vizzari

With the general objective of producing a 2018–2020 Land Use/Land Cover (LULC) map of the Maiella National Park (central Italy), useful for a future long-term LULC change analysis, this research aimed to develop a Landsat 8 (L8) data composition and classification process using Google Earth Engine (GEE). In this process, we compared two pixel-based (PB) and two object-based (OB) approaches, assessing the advantages of integrating the textural information in the PB approach. Moreover, we tested the possibility of using the L8 panchromatic band to improve the segmentation step and the object’s textural analysis of the OB approach and produce a 15-m resolution LULC map. After selecting the best time window of the year to compose the base data cube, we applied a cloud-filtering and a topography-correction process on the 32 available L8 surface reflectance images. On this basis, we calculated five spectral indices, some of them on an interannual basis, to account for vegetation seasonality. We added an elevation, an aspect, a slope layer, and the 2018 CORINE Land Cover classification layer to improve the available information. We applied the Gray-Level Co-Occurrence Matrix (GLCM) algorithm to calculate the image’s textural information and, in the OB approaches, the Simple Non-Iterative Clustering (SNIC) algorithm for the image segmentation step. We performed an initial RF optimization process finding the optimal number of decision trees through out-of-bag error analysis. We randomly distributed 1200 ground truth points and used 70% to train the RF classifier and 30% for the validation phase. This subdivision was randomly and recursively redefined to evaluate the performance of the tested approaches more robustly. The OB approaches performed better than the PB ones when using the 15 m L8 panchromatic band, while the addition of textural information did not improve the PB approach. Using the panchromatic band within an OB approach, we produced a detailed, 15-m resolution LULC map of the study area.


Sign in / Sign up

Export Citation Format

Share Document