SoilGrids 2.0: producing soil information for the globe with quantified spatial uncertainty

SOIL, Journal Year: 2021, Volume and Issue: 7(1), P. 217 - 240

Published: June 14, 2021

Abstract. SoilGrids produces maps of soil properties for the entire globe at medium spatial resolution (250 m cell size) using state-of-the-art machine learning methods to generate necessary models. It takes as inputs observations from about 240 000 locations worldwide and over 400 global environmental covariates describing vegetation, terrain morphology, climate, geology hydrology. The aim this work was production properties, with cross-validation, hyper-parameter selection quantification spatially explicit uncertainty, implemented in version 2.0 product incorporating practices adapting them digital mapping legacy data. paper presents evaluation predictions produced organic carbon content, total nitrogen, coarse fragments, pH (water), cation exchange capacity, bulk density texture fractions six standard depths (up 200 cm). quantitative showed metrics line previous global, continental large-region studies. qualitative that coarse-scale patterns are well reproduced. uncertainty scale highlighted need more observations, especially high-latitude regions.

Language: Английский

---

Machine learning and deep learning—A review for ecologists

Methods in Ecology and Evolution, Journal Year: 2023, Volume and Issue: 14(4), P. 994 - 1016

Published: Feb. 13, 2023

Abstract The popularity of machine learning (ML), deep (DL) and artificial intelligence (AI) has risen sharply in recent years. Despite this spike popularity, the inner workings ML DL algorithms are often perceived as opaque, their relationship to classical data analysis tools remains debated. Although it is assumed that excel primarily at making predictions, can also be used for analytical tasks traditionally addressed with statistical models. Moreover, most discussions reviews on focus mainly DL, failing synthesise wealth different advantages general principles. Here, we provide a comprehensive overview field starting by summarizing its historical developments, existing algorithm families, differences traditional tools, universal We then discuss why when models prediction where they could offer alternatives methods inference, highlighting current emerging applications ecological problems. Finally, summarize trends such scientific causal ML, explainable AI, responsible AI may significantly impact future. conclude powerful new predictive modelling analysis. superior performance compared explained higher flexibility automatic data‐dependent complexity optimization. However, use inference still disputed predictions creates challenges interpretation these Nevertheless, expect become an indispensable tool ecology evolution, comparable other tools.

Language: Английский

---

African soil properties and nutrients mapped at 30 m spatial resolution using two-scale ensemble machine learning

Scientific Reports, Journal Year: 2021, Volume and Issue: 11(1)

Published: March 17, 2021

Soil property and class maps for the continent of Africa were so far only available at very generalised scales, with many countries not mapped all. Thanks to an increasing quantity availability soil samples collected field point locations by various government and/or NGO funded projects, it is now possible produce detailed pan-African nutrients, including micro-nutrients fine spatial resolutions. In this paper we describe production a 30 m resolution Information System African using, date, most comprehensive compilation ([Formula: see text]) Earth Observation data. We produced predictions pH, organic carbon (C) total nitrogen (N), carbon, effective Cation Exchange Capacity (eCEC), extractable-phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), sodium (Na), iron (Fe), zinc (Zn)-silt, clay sand, stone content, bulk density depth bedrock, three depths (0, 20 50 cm) using 2-scale 3D Ensemble Machine Learning framework implemented in mlr (Machine R) package. As covariate layers used 250 (MODIS, PROBA-V SM2RAIN products), (Sentinel-2, Landsat DTM derivatives) images. Our fivefold Cross-Validation results showed varying accuracy levels ranging from best performing pH (CCC = 0.900) more poorly predictable extractable phosphorus 0.654) sulphur 0.708) bedrock. Sentinel-2 bands SWIR (B11, B12), NIR (B09, B8A), bands, vertical derived DTM, overall important covariates. Climatic data images-SM2RAIN, bioclimatic variables MODIS Land Surface Temperature-however, remained as predicting chemical continental scale. This publicly 30-m aims supporting numerous applications, fertilizer policies investments, agronomic advice close yield gaps, environmental programs, or targeting nutrition interventions.

Language: Английский

---

Machine learning-based global maps of ecological variables and the challenge of assessing them

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: April 22, 2022

The recent wave of published global maps ecological variables has caused as much excitement it received criticism. Here we look into the data and methods mostly used for creating these maps, discuss whether quality predicted values can be assessed, globally locally.

Language: Английский

---

FLUXNET-CH<sub>4</sub>: a global, multi-ecosystem dataset and analysis of methane seasonality from freshwater wetlands

Earth system science data, Journal Year: 2021, Volume and Issue: 13(7), P. 3607 - 3689

Published: July 29, 2021

Abstract. Methane (CH4) emissions from natural landscapes constitute roughly half of global CH4 contributions to the atmosphere, yet large uncertainties remain in absolute magnitude and seasonality emission quantities drivers. Eddy covariance (EC) measurements flux are ideal for constraining ecosystem-scale due quasi-continuous high-temporal-resolution measurements, coincident carbon dioxide, water, energy lack ecosystem disturbance, increased availability datasets over last decade. Here, we (1) describe newly published dataset, FLUXNET-CH4 Version 1.0, first open-source dataset EC (available at https://fluxnet.org/data/fluxnet-ch4-community-product/, access: 7 April 2021). includes half-hourly daily gap-filled non-gap-filled aggregated fluxes meteorological data 79 sites globally: 42 freshwater wetlands, 6 brackish saline formerly drained ecosystems, rice paddy sites, 2 lakes, 15 uplands. Then, (2) evaluate representativeness wetland coverage globally because majority 1.0 wetlands which a substantial source total atmospheric emissions; (3) provide estimates seasonal variability predictors fluxes. Our analysis suggests that cover bioclimatic attributes (encompassing energy, moisture, vegetation-related parameters) arctic, boreal, temperate regions but only sparsely humid tropical regions. Seasonality metrics vary considerably across latitudinal bands. In (except those between 20∘ S N) spring onset elevated starts 3 d earlier, season lasts 4 longer, each degree Celsius increase mean annual air temperature. On average, increasing lags behind soil warming by 1 month, with very few experiencing prior warming. contrast, these experience rising gross primary productivity (GPP). The timing peak summer does not correlate either temperature or GPP. results parameters modeling highlight cannot be predicted GPP (i.e., peak). is powerful new resource diagnosing understanding role terrestrial ecosystems climate drivers cycle, future additions site years collection will added value this database. All available https://doi.org/10.5281/zenodo.4672601 (Delwiche et al., Additionally, raw used extract can downloaded https://fluxnet.org/data/fluxnet-ch4-community-product/ (last 2021), complete list individual DOIs provided Table paper.

Language: Английский

---

Expanding agroforestry can increase nitrate retention and mitigate the global impact of a leaky nitrogen cycle in croplands

Nature Food, Journal Year: 2022, Volume and Issue: 4(1), P. 109 - 121

Published: Dec. 28, 2022

The internal soil nitrogen (N) cycle supplies N to plants and microorganisms but may induce pollution in the environment. Understanding variability of gross cycling rates resulting from global spatial heterogeneity climatic edaphic variables is essential for estimating potential risk loss. Here we compiled 4,032 observations 398 published

Language: Английский

---

Spatially autocorrelated training and validation samples inflate performance assessment of convolutional neural networks

ISPRS Open Journal of Photogrammetry and Remote Sensing, Journal Year: 2022, Volume and Issue: 5, P. 100018 - 100018

Published: June 21, 2022

Deep learning and particularly Convolutional Neural Networks (CNN) in concert with remote sensing are becoming standard analytical tools the geosciences. A series of studies has presented seemingly outstanding performance CNN for predictive modelling. However, such models is commonly estimated using random cross-validation, which does not account spatial autocorrelation between training validation data. Independent method, dependence will inevitably inflate model performance. This problem ignored most CNN-related suggests a flaw their procedure. Here, we demonstrate how neglecting during cross-validation leads to an optimistic assessment, example tree species segmentation multiple, spatially distributed drone image acquisitions. We evaluated CNN-based predictions test data sampled from 1) randomly hold-outs 2) blocked hold-outs. Assuming that block provides realistic performance, holdouts overestimated by up 28%. Smaller sample size increased this optimism. Spatial among observations was significantly higher within than different Thus, should be tested strategies multiple independent Otherwise, any geospatial deep method likely overestimated.

Language: Английский

---

Drivers and trends of global soil microbial carbon over two decades

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: July 20, 2022

Abstract Soil microorganisms are central to sustain soil functions and services, like carbon nutrient cycling. Currently, we only have a limited understanding of the spatial-temporal dynamics microorganisms, restricting our ability assess long-term effects climate land-cover change on microbial roles in biogeochemistry. This study assesses temporal trends biomass identifies main drivers regionally globally detect areas sensitive these environmental factors. Here, combined global data set, random forest modelling, layers predict stocks from 1992 2013. decreased by 3.4 ± 3.0% (mean 95% CI) between 2013 for predictable regions, equivalent 149 Mt being lost over period, or ~1‰ C. Northern with high experienced strongest decrease, mostly driven increasing temperatures. In contrast, was weaker driver carbon, but had, some cases, important regional effects.

Language: Английский

---

Dealing with clustered samples for assessing map accuracy by cross-validation

Ecological Informatics, Journal Year: 2022, Volume and Issue: 69, P. 101665 - 101665

Published: May 5, 2022

Mapping of environmental variables often relies on map accuracy assessment through cross-validation with the data used for calibrating underlying mapping model. When points are spatially clustered, conventional leads to optimistically biased estimates accuracy. Several papers have promoted spatial as a means tackle this over-optimism. Many these blame autocorrelation cause bias and propagate widespread misconception that proximity calibration validation invalidates classical statistical maps. We present evaluate alternative approaches assessing from clustered sample data. The first method uses inverse sampling-intensity weighting correct selection bias. Sampling-intensity is estimated by two-dimensional kernel approach. two other model-based methods rooted in geostatistics, where assumes homogeneity residual variance over study area whilst second accounts heteroscedasticity function sampling intensity. were tested compared against k-fold blocked estimate metrics above-ground biomass soil organic carbon stock maps covering western Europe. Results acquired 100 realizations five designs ranging non-clustered strongly confirmed heteroscedastic had smaller than all but most design. For design large portions predicted extrapolation, was closest reference metrics, still biased. such cases, extrapolation best avoided additional or limitation prediction area. Weighted recommended moderately samples, while random suits fairly regularly spread samples.

Language: Английский

---

Assessing and improving the transferability of current global spatial prediction models

Global Ecology and Biogeography, Journal Year: 2023, Volume and Issue: 32(3), P. 356 - 368

Published: Jan. 26, 2023

Abstract Aim Global‐scale maps of the environment are an important source information for researchers and decision makers. Often, these created by training machine learning algorithms on field‐sampled reference data using remote sensing as predictors. Since field samples often sparse clustered in geographic space, model prediction requires a transfer trained to regions where no available. However, recent studies question feasibility predictions far beyond location data. Innovation We propose novel workflow spatial predictive mapping that leverages developments this combines them innovative ways with aim improved transferability performance assessment. demonstrate, evaluate discuss from recently published global environmental maps. Main conclusions Reducing predictors those relevant leads increase map accuracy without decrease quality areas high sampling density. Still, reliable gap‐free were not possible, highlighting their evaluation hampered limited availability

Language: Английский

---