A Machine Learning (kNN) Approach to Predicting Global Seafloor Total Organic Carbon

Lee, Taylor; Wood, W. T.; Phrampus, B. J.

doi:10.1029/2018gb005992

Cited by 105 publications

(156 citation statements)

References 21 publications

(36 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous studies have predicted OC concentrations and sediment porosity separately to calculate OC stocks (Diesing et al, 2017;Lee et al, 2019;Wilson et al, 2018). Here, we first calculate OC density from concurrent measurements of OC concentrations and sediment dry bulk densities or porosities.…”

Section: Organic Carbon Densitymentioning

confidence: 99%

Organic carbon in surface sediments of the North Sea and Skagerrak

Diesing

Thorsnes

Bjarnadóttir

2020

Preprint

View full text Add to dashboard Cite

Abstract. Continental shelf sediments are places of both rapid organic carbon turnover and accumulation, while at the same time increasingly subjected to human-induced disturbances. Recent research suggests that shelf sediments might have a role to play as a natural climate solution, e.g. by protecting the seafloor against human-induced disturbance. However, we have an incomplete understanding about the centres of organic carbon accumulation and storage on continental shelves. To better constrain the rate of accumulation and the mass of organic carbon that is stored in sediments, we developed and applied a spatial modelling framework that allows to estimate those quantities from sparse observations and predictor variables known or suspected to influence the spatial patterns of these parameters. This paper presents spatial distribution patterns of organic carbon densities and accumulation rates in the North Sea and Skagerrak. We found that organic carbon stocks and accumulation rates are highest in the Norwegian Trough, while large parts of the North Sea are characterised by low stocks and zero net-accumulation. The total stock of organic carbon that is stored in the upper 0.1 m of sediments amounted to 230.5 ± 134.5 Tg, of which approximately 26 % are stored in the Norwegian Trough. Rates of organic carbon accumulation in the Norwegian Trough are on par with those reported from nearby fjords. We provide baseline datasets that could be used in marine management, e.g. for the establishment of carbon protection zones. Additionally, we highlight the complex nature of continental shelves with zones of rapid carbon cycling and accumulation juxtaposed, which will require further detailed and spatially explicit analyses to constrain sedimentary organic carbon stocks and accumulation rates globally.

show abstract

Section: Organic Carbon Densitymentioning

confidence: 99%

Organic carbon in surface sediments of the North Sea and Skagerrak

Diesing

Thorsnes

Bjarnadóttir

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…A 10-fold cross validation (CV), an approach frequently used in model validation (e.g. Li et al, 2011;Gasch et al, 2015;Ließ et al, 2016;Meyer et al, 2016bMeyer et al, , 2018, was applied to validate the RF model. CV involved partitioning the dataset into 10 equally sized folds.…”

Section: Model Validationmentioning

confidence: 99%

“…Studies comparing RF against other machine learning algorithms reported different trends: in some cases, RF was superior in terms of prediction performance (e.g. Li et al, 2011;Cracknell and Reading, 2014), whereas in other cases, no strong differences were observed between the different methods (e.g. Goetz et al, 2015;Meyer et al, 2016b).…”

Section: Suggestions For Further Improvement Of Performancementioning

confidence: 99%

“…In geosciences and remote sensing, machine learning methods have been used to predict soil properties (Gasch et al, 2015;Ließ et al, 2016;Meyer et al, 2018), air temperatures (Meyer et al, 2016a(Meyer et al, , 2018, biomass (Meyer et al, 2017), and the elasticity modulus of granitic rocks (Karakus, 2011). Applications also extended into marine settings, involving the prediction of sediment mud content off southwest Australia (Li et al, 2011), as well as parameters such as seafloor porosity (Martin et al, 2015;Wood et al, 2018), seafloor biomass (Wei et al, 2010), and seafloor total organic carbon (Wood et al, 2018;Lee et al, 2019) on a global scale. These studies were in general restricted to the prediction of one value per geographic location; the prediction of multiple values, such as depth profiles, has, to our knowledge, not been attempted before.…”

Section: Introductionmentioning

confidence: 99%

“…RF has been repeatedly found to be superior to other machine learning methods. For example, Li et al (2011) tested 23 machine learning algorithms -including RF, SVM, and kriging methods -to predict mud content in marine sediments and found that RF, along with RF combined with ordinary kriging or inverse distance squared, provided the best prediction results. Cracknell and Reading (2014) applied five machine learning methods to lithology classification of multispectral satellite data and reported higher classification accuracy for RF than for naive Bayes, SVM, ANN, and k-nearest neighbours.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Prediction of seismic P-wave velocity using machine learning

Dumke

Berndt

2019

Solid Earth

View full text Add to dashboard Cite

Abstract. Measurements of seismic velocity as a function of depth are generally restricted to borehole locations and are therefore sparse in the world's oceans. Consequently, in the absence of measurements or suitable seismic data, studies requiring knowledge of seismic velocities often obtain these from simple empirical relationships. However, empirically derived velocities may be inaccurate, as they are typically limited to certain geological settings, and other parameters potentially influencing seismic velocities, such as depth to basement, crustal age, or heat flow, are not taken into account. Here, we present a machine learning approach to predict the overall trend of seismic P-wave velocity (vp) as a function of depth (z) for any marine location. Based on a training dataset consisting of vp(z) data from 333 boreholes and 38 geological and spatial predictors obtained from publicly available global datasets, a prediction model was created using the random forests method. In 60 % of the tested locations, the predicted seismic velocities were superior to those calculated empirically. The results indicate a promising potential for global prediction of vp(z) data, which will allow the improvement of geophysical models in areas lacking first-hand velocity data.

show abstract

Towards time and cost‐efficient habitat assessment: Challenges and opportunities for benthic ecology and management

Schenone,

Azhar,

Delmas

et al. 2023

Aquatic Conservation

View full text Add to dashboard Cite

To the untrained eye, soft‐sediment habitats are often viewed as vast expanses of featureless sand and mud. With new technology, it is finally possible to begin to map the ecological heterogeneity of these ecosystems to reveal their importance for ecosystem service provision and open up new opportunities to manage their use and protect them. When compared with terrestrial or marine vegetated habitats, the characterization of unvegetated soft sediments still lacks rapid and cost‐effective practices that allow the direct collection of ecological data at multiple spatial and temporal scales facilitating direct relevance to conservation and management. New technologies that allow for the building of maps that incorporate multiple scales of heterogeneity also provide new ways of upscaling necessarily detailed and small‐scale process studies. The study of the relationship between benthic structure, biodiversity and ecosystem functioning can benefit from a range of established and novel techniques and creates new opportunities to empirically test generalities and predictions based on theory or small‐scale experimental studies. The capacity to map ecological changes rapidly and at scale and to link small‐scale experiments to landscapes uncovers new insights for seafloor conservation including rapid response to disturbance events, broad‐scale ecologically meaningful mapping and assessments of biodiversity–ecosystem function and biodiversity–ecosystem services relationships. As technology advances and becomes more accessible to all, it offers promising new avenues of research to expand our current capabilities, and the adoption of nested approaches, with a synergy of different samplings simultaneously covering several scales and resolutions, is likely to have the most benefit, supporting decision‐making with ecologically insightful data, as well as the timely assessment of the outcomes of conservation actions.

show abstract

A Machine Learning (kNN) Approach to Predicting Global Seafloor Total Organic Carbon

Cited by 105 publications

References 21 publications

Organic carbon in surface sediments of the North Sea and Skagerrak

Organic carbon in surface sediments of the North Sea and Skagerrak

Prediction of seismic P-wave velocity using machine learning

Towards time and cost‐efficient habitat assessment: Challenges and opportunities for benthic ecology and management

Contact Info

Product

Resources

About