Search

The weakness of global metrics for assessing the accuracy of satellite-derived bathymetry (SDB) is demonstrated and the magnitude of local accuracy variability quantified. SDB was developed for three small areas (500 m-by-1500 m) near Key West Florida (United States) that varied in depth and geomorphometric complexity. Data employed were 2016 and 2019 airborne lidar and three 2016 and four 2019 Landsat images. The modelling methodology was a widely used quasi-empirical linear regression-based approach. Local accuracy of SDB depth change – which was expected to be pronounced between 2016 and 2019 due to the occurrence of Category 5 Hurricane Irma in 2017 – was also calculated and examined. The global correlation (R2) between lidar-based depth and SDB for 2016 and 2019 varied from about 0.1 to 0.9 which produced relative (to depth) root mean square errors (RMSEs) between 5% and 25%. No image was consistently best for all three areas nor for either 2016 and 2019. SDB-based depth change was poorly correlated to the reference lidar-based depth change for all three areas. Spatial depiction of SDB accuracy showed considerable variability even within the three relatively small areas employed for 2016 and 2019 SDB depth estimates and 2016–2019 SDB depth change estimates. [ABSTRACT FROM AUTHOR]

Color is an important cue in object recognition and classification problems. In underwater imagery, colors undergo strong distortion due to light propagation through an absorbing and scattering medium. Distortions depend on a number of complex phenomena, the most important being wavelength-dependent absorption and the sensitivity of sensors in trichromatic cameras. It has been shown previously that unique reconstruction in this case is not possible—at least for a simplified image formation model. In this paper, the authors use numerical simulations to demonstrate that this statement also holds for the underwater image formation model that is currently the most sophisticated. [ABSTRACT FROM AUTHOR]

Most work to date on satellite-derived bathymetry (SDB) depth change estimates water depth at individual times t1 and t2 using two separate models and then differences the model estimates. An alternative approach is explored in this study: a multi-temporal Sentinel-2 image is created by "stacking" the bands of the times t1 and t2 images, geographically coincident reference data for times t1 and t2 allow for "true" depth change to be calculated for the pixels of the multi-temporal image, and this information is used to fit a single model that estimates depth change directly rather than indirectly as in the model-differencing approach. The multi-temporal image approach reduced the depth change RMSE by about 30%. The machine learning modelling method (categorical boosting) outperformed linear regression. Overfitting of models was limited even for the CatBoost models having the maximum number of variables examined. The visible Sentinel-2 spectral bands contributed most to the model predictions. Though the multi-temporal stacked image approach produced clearly superior depth change estimates compared to the conventional approach, it is limited only to those areas for which geographically coincident multi-temporal reference/"true" depth data exist. [ABSTRACT FROM AUTHOR]

The oceans remain one of Earth's most enigmatic frontiers, with approximately 75% of the world's oceans still unmapped. To create a seamless digital bathymetric model (DBM) from sparse bathymetric datasets, interpolation is employed, but this introduces uncertainties of unknown nature. This study aims to estimate and characterize these uncertainties, which is important in many fields, particularly nautical charting, and navigational safety. Complete seafloor coverage sonar depth data for five testbeds that varied in slope and roughness are sampled at a range of densities (1% to 50%) and interpolated across an entire area using Spline, Inverse Distance Weighting (IDW), and Linear interpolation. The resulting uncertainties are evaluated from both scientific and operational perspectives. Employing linear regression and machine learning techniques, the relationships between these uncertainties and ancillary parameters (distance to the nearest measurement, seabed roughness, and slope) are examined for quantifying and characterizing the interpolation uncertainty. Results indicate insignificant operational differences among the three interpolators in depth estimation, as well as the statistical significance of the examined uncertainty predictors. Additionally, findings suggest the potential presence of unaccounted-for factors shaping uncertainty, yet this work lays a foundational understanding for improving the estimate of uncertainty in DBMs. [ABSTRACT FROM AUTHOR]

The goal of this work was to evaluate if routinely collected but seldom used airborne lidar metadata – 'point attribute data' (PAD) – analyzed using machine learning/artificial intelligence can improve extraction of shallow-water (less than 20 m) bathymetry from lidar point clouds. Extreme gradient boosting (XGB) models relating PAD to an existing bathymetry/not bathymetry classification were fitted and evaluated for four areas near the Florida Keys. The PAD examined include 'pulse specific' information such as the return intensity and PAD describing flight path consistency. The R2 values for the XGB models were between 0.34 and 0.74. Global classification accuracies were above 80% although this reflected a sometimes extreme Bathy/NotBathy imbalance that inflated global accuracy. This imbalance was mitigated by employing a probability decision threshold (PDT) that equalizes the true positive (Bathy) and true negative (NotBathy) rates. It was concluded that 1) the strength of the bathymetric signal in the PAD should be sufficient to increase accuracy of density-based lidar point cloud bathymetry extraction methods and 2) ML can successfully model the relationship between the PAD and the Bathy/NotBathy classification. A method is also presented to examine the spatial and feature-space distribution of errors that will facilitate quality assurance and continuous improvement. [ABSTRACT FROM AUTHOR]

To automate extraction of bathymetric soundings from lidar point clouds, two machine learning (ML1) techniques were combined with a more conventional density-based algorithm. The study area was four data "tiles" near the Florida Keys. The density-based algorithm determined the most likely depth (MLD) for a grid of "estimation nodes" (ENs). Unsupervised k-means clustering determined which EN's MLD depth and associated soundings represented ocean depth rather than ocean surface or noise to produce a preliminary classification. An extreme gradient boosting (XGB) model was fitted to pulse return metadata – e.g. return intensity, incidence angle – to produce a final Bathy/NotBathy classification. Compared to an operationally produced reference classification, the XGB model increased global accuracy and decreased the false negative rate (FNR) – i.e. undetected bathymetry – that are most important for nautical navigation for all but one tile. Agreement between the final XGB and operational reference classifications ranged from 0.84 to 0.999. Imbalance between Bathy and NotBathy was addressed using a probability decision threshold that equalizes the FNR and the true positive rate (TPR). Two methods are presented for visually evaluating differences between the two classifications spatially and in feature-space. [ABSTRACT FROM AUTHOR]

The impact of multibeam echosounder (MBES) operations on marine mammals has been less studied compared to military sonars. To contribute to the growing body of MBES knowledge, echolocation clicks of foraging Cuvier's beaked whales were detected on the Southern California Antisubmarine Warfare Range (SOAR) hydrophones during two MBES surveys and assembled into foraging events called group vocal periods (GVPs). Four GVP characteristics were analyzed Before, During, and After 12 kHz MBES surveys at the SOAR in 2017 and 2019 to assess differences in foraging behavior with respect to the mapping activity. The number of GVP per hour increased During and After MBES surveys compared with Before. There were no other differences between non-MBES and MBES periods for the three other characteristics: the number of clicks per GVP, GVP duration, and click rate. These results indicate that there was not a consistent change in foraging behavior during the MBES surveys that would suggest a clear response. The animals did not leave the range nor stop foraging during MBES activity. These results are in stark contrast to those of analogous studies assessing the effect of Naval mid-frequency active sonar on beaked whale foraging, where beaked whales stopped echolocating and left the area. [ABSTRACT FROM AUTHOR]

To identify potentially critical changes in an area's ongoing ability to produce multiple ecosystem services, a monitoring system was designed and implemented for the Mornington Peninsula and Westernport Biosphere Reserve in southeast Australia. The system is underpinned by an 'environmental vital signs' (EVS) approach that was adopted to provide early warning of critical changes in human and natural characteristics of the area. The six themes monitored are non-coastal water, land including vegetation, biodiversity, natural heritage, built environment (including human population and economic activity), and coasts. These are monitored for the entire area, and each of its five constituent town council areas. After a critical change in any of these is identified, further investigation is required to identify causal factors and, if required, determine an appropriate response. The system relies on data available from external (third-party) organisations to monitor the natural and human characteristics of the area that were important in its designation as a UNESCO biosphere reserve. Strengths and weaknesses associated with the use of third-party data are discussed. These include adoption of baseline years and data reporting periods for different factors, costs, and data quality. [ABSTRACT FROM AUTHOR]

This paper evaluates the opportunity provided by global interferometric radar datasets for monitoring deforestation, degradation and forest regrowth in tropical and semi-arid environments. The paper describes an easy to implement method for detecting forest spatial changes and estimating their magnitude. The datasets were acquired within space-borne high spatial resolutions radar missions at near-global scales thus being significant for monitoring systems developed under the United Framework Convention on Climate Change (UNFCCC). The approach presented in this paper was tested in two areas located in Indonesia and Australia. Forest change estimation was based on differences between a reference dataset acquired in February 2000 by the Shuttle Radar Topography Mission (SRTM) and TanDEM-X mission (TDM) datasets acquired in 2011 and 2013. The synergy between SRTM and TDM datasets allowed not only identifying changes in forest extent but also estimating their magnitude with respect to the reference through variations in forest height. [ABSTRACT FROM AUTHOR]

The backscatter predicted by three common surface scattering models (the Integral Equation Model (IEM), the Dubois, and the Oh models) was evaluated against fully polarized L-band airborne observations. Before any site-specific calibration, the Oh model was found to be the most accurate among the three, with mean errors between the simulated and the observed backscatter of 1.2 dB ( ±2.6 dB standard deviation of the error) and -0.4 dB ( ±2.4 dB) for HH and VV polarizations, respectively, while the IEM and Dubois presented larger errors, with a maximum of 4.5 dB ( ±2 dB) for the IEM in VV polarization. The backscatter errors were observed to be related to surface roughness, another major factor determining the electromagnetic scattering at the soil surface. An existing semiempirical calibration of the surface roughness correlation length was therefore applied to improve the mismatch between modeled and observed backscatters. The application of the semiempirical calibration led to a significant improvement of the backscatter prediction for the IEM. After calibration, the IEM outperformed the Oh model, resulting in a mean backscatter error of -0.3 dB ( ±1.1 dB) and -0.2 ( ±1.2 dB) for HH and VV polarizations, respectively. To test the robustness of the semiempirical calibration, calibration functions derived from an independent data set were applied and shown to also improve the (uncalibrated) IEM performance, suggesting that the calibration procedure is relatively robust for global application. [ABSTRACT FROM PUBLISHER]

This paper investigated the effectiveness of frequently used parametric and nonparametric models for biomass retrieval from L-band radar backscatter. Two areas, one in Spain and one in Australia, characterized by different tree species, forest structure, and field sampling designs were selected to demonstrate that retrieval error metrics are similar for different local conditions and sampling characteristics. A mixed-model retrieval strategy was proposed to reduce the overall (i.e., across the entire biomass range) as well as by-biomass-interval errors. Significant relationships were found between aboveground biomass and radar backscatter with most of the backscatter dynamic range being limited to a fairly low range of biomass values ( t/ha) in both study areas. Biomass retrieval errors were largely similar for all parametric and nonparametric models tested. However, some parametric models consistently provided lower correlation between the observed and the predicted biomass while nonparametric models generally provided an unbiased estimation. A mixed-model retrieval strategy was shown to reduce biomass estimation errors by up to 15%. Biomass retrieval errors were highly variable within the L-band sensitivity interval, suggesting that overall accuracy estimates should be used with care, particularly for low biomass intervals ( t/ha) where surface scattering could dominate the total backscatter. Despite exhibiting the highest dynamic range, low biomass areas were characterized by the highest estimation errors (in excess of 80%). Conversely, relative estimation errors were as low as 20%-35% for the 30-75 t/ha biomass intervals, while at higher biomass levels, the estimation error increased due to signal saturation. [ABSTRACT FROM PUBLISHER]

This letter evaluates the biomass-retrieval error in pine-dominated stands when using high-spatial-resolution airborne measurements from fully polarimetric L-band radar and airborne laser scanning sensors. Information on total above-ground biomass was estimated through allometric relationships from plot-level field measurements. Multiple-linear-regression models were developed to model relationships between biomass and radar/lidar data. Overall, lidar data provided lower estimation errors (17.2 \t\cdot \ha^-1, 28% relative) when compared with radar data (30.3 \t\cdot\ha^-1, 61% relative). However, for the 30–100 \t\cdot\ha^-1 biomass range, the relative error from radar-based models was only 9% higher than that from lidar-based models. This suggests that high-spatial-resolution radar data could provide fundamentally similar results to lidar for some biomass intervals. This is an important finding for large-scale biomass estimation that needs to rely upon satellite data, as there are no lidar satellites planned for the foreseeable future. [ABSTRACT FROM PUBLISHER]

In response to disparate advances in delivering spatial information to support agricultural extension activities, the Extension Activity Support System (EASY) project was established to develop a vision statement and conceptual design for such a system based on a national needs assessment. Personnel from across Australia were consulted and a review of existing farm information/management software undertaken to ensure that any system that is eventually produced from the EASY vision will build on the strengths of existing efforts. This paper reports on the collaborative consultative process undertaken to create the EASY vision as well as the conceptual technical design and business models that could support a fully functional spatially enabled online system. [ABSTRACT FROM AUTHOR]

A decision framework called Deliberative Multi-Criteria Evaluation (DMCE) was developed and deployed to prioritize biosecurity risks using a variety of subjective and objective information. To aid stakeholders in the prioritization of Emergency Plant Pest (EPP) species risk, we presented them with outputs from a Stella-based bio-economic pest risk model, and probable ecological and socioeconomic impacts. The stakeholder participants weighted the consequence criteria they deemed to have the highest expected impact. The methodology featured an uncertain set of parameters, multiple iterations of criteria weighting along with real-time sensitivity analysis. Of the five criteria, economic cost was weighted the highest at 26% while landscape amenity was weighted the lowest at the 10-12% range. The increased understanding and support gained by stakeholders through the DMCE process provided a greater likelihood of the sanctioning of the policies concerned and progress toward desired outcomes. [ABSTRACT FROM AUTHOR]

Bushfire threat was evaluated for built structures for three areas in Victoria (Australia) that had been impacted by the devastating 16 February 1983 Ash Wednesday fires. Structures were mapped for 1982 and 2006 using human inter- pretation of high-resolution (0.35-m pixels) digital orthophotographs. Damage to structures from the 1983 Ash Wednesday fires was also evaluated using human interpretation of the digital orthophotographs. Approximately 25% of the structures present were not mapped due to either interpreter error or overhanging vegetation. The majority of unmapped structures were sheds and garages. The error of omission for houses was between 7 and 10% with the error of commission for houses being less than 0.5%. Bushfire threat was modelled using information about topographic slope and aspect, forest vegetation, and prevailing wind direction during days of high fire danger. The method detected a substantial change in bushfire threat from 1982 to 2006 for one of the three study sites whereas no change in overall bushfire threat was observed for the other two. Considering the location of structures built since 1982, these results appeared reasonable. However, the 1982 bushfire threat was not related to actual structure damage sustained during the 1983 Ash Wednesday fires. Estimating bushfire threat using this methodology cost AU$6 per structure or AU$4.60 per property. [ABSTRACT FROM AUTHOR]

A fuzzy methodology is presented for evaluating the reliability of satellite imagery-derived, continent-wide (Australia) land-cover deforestation/regrowth maps covering the period of 1972 to 2000 in ten discrete time periods. The methodology uses aerial photographs as its reference data and accommodates the difficulty inherent in determining definitively from an aerial photograph, whether a sample point is Forest or Non-forest by permitting interpreters to identify their level of certainty, i.e., Definitely Forest, Probably Forest, Uncertain, Probably Non-forest, or Definitely Non- forest. This information is then cross-tabulated against the Forest/Non-forest classification for the classified image closest in date to the photo date. Information from several photographs is summarized over a larger geographic area and over all time periods. Subsequently, temporal lineage information for each sample pixel is extracted from the 1972 to 2000 series of classified images to determine if a pixel's lineage is Forest Throughout, Non-forest Throughout, Deforestation, Regrowth, or Cyclic. The fuzzy evaluation for individual pixels is then tabulated against this lineage information to identify if pixels of any particular lineage have an elevated tendency to be misclassified. The methodology provides a means by which problems in the map production methodology can be improved as future time slices are added. [ABSTRACT FROM AUTHOR]

A 100-m-by-100-m (1 ha) square study area was established in the boreal forest of Quebec, Canada. The area was selected because of the presence of roughly equal amounts of three distinct forest types as identified from aerial photographs through subjective human interpretation. Within the study area, the position of every tree having a diameter breast height (dbh) of 9.1 cm and larger was determined in x- and y-coordinates. Sample units of various sizes and types were then established at a variety of spacings within the study area and the forest volume determined for each. A variety of analyses was conducted on each spacing/sample unit size and type combination to assess the nature of forest volume across cartographic boundaries. It was determined that at closer spacings and/or in using larger sample units, forest volume behaves as a continuum. However, the nature of the change in volume across forest stand boundaries was not (as was expected a priori) a gradual decrease in volume from a high volume stand to a low volume stand with relatively homogeneous areas further from the boundary. Instead it was found that, for a forest type boundary having stands of different densities but the same species on either side, the boundary was an area of relatively high volume with volume decreasing on both sides of the boundary as the distance from the boundary increased. Regardless of density, for a forest type boundary having stands of different species on either side, the boundary was an area of relatively low volume with volume increasing on either side of the boundary as the distance from the boundary increased. Readers are cautioned that these results were extracted from a single study site and that generalizing them to other forest conditions should only be done with care. [ABSTRACT FROM AUTHOR]

Focuses on forest productivity models as a function of digital ecophysiographic data and distance from a mapped boundary in Victoria, Australia. Relationship between digital spatial data and volume; Influence of the developed equation on the confidence interval width for log volume; Variables describing forest conditions.