Pedometrics 2024: Addressing the 10 Challenges
Feb 5 – 9, 2024
New Mexico State University
Las Cruces, NM USA
Pre-conference workshops, field trips, active social programs, and post-conference excursions are being planned!
The 10 challenges:
Pedometrics has broadened its scope over the past two decades. During this time, a general demand for quantitative digital soil information for environmental modelling and management has compelled pedometricians to address many soil-related questions from a quantitative point of view.
While scientific progress is largely an autonomous process that is difficult to steer, research efforts could benefit from an agenda with pressing pedometric research topics. This conference is focused around ten recent or longstanding pedometrics challenges (Wadoux et al, 2021). The ten challenges were selected through a collaborative effort and will serve to help organize conference sessions to foster collaboration among soil scientists.
Alexandre M.J.-C. Wadoux, Gerard B.M. Heuvelink, R. Murray Lark, Philippe Lagacherie, Johan Bouma, Vera L. Mulder, Zamir Libohova, Lin Yang, Alex B. McBratney. Ten challenges for the future of pedometrics, Geoderma,
Volume 401, 2021. https://doi.org/10.1016/j.geoderma.2021.115155.
The 10 challenges are as follows.
|How can we better understand soil formation?|
|1||Can we produce quantitative models of the complex short and long-term processes of soil formation which are predictive of the spatio-temporal variation of soil properties?||Soil change ; Forecast ; Quantitative models of pedogenetic processes ; Time series ; Dynamic mechanistic models ; Soil-landscape evolution models ; Quantifying soil genesis|
|2||Can we develop a quantitative and numerical global soil classification that unifies the existing systems and enables transfer between them?||Numerical soil classification ; Soil taxonomy ; Local and regional applications ; Communication between classification systems ; Similarities between soil profiles ; Translation between systems ; Near real-time soil classification|
|3||In what ways can we use data-driven models to learn about pedological processes?||Interpretation of complex models ; Multi-scale drivers of soil variation ; Functional relationships between covariates and soil data ; Hypothesis discovery ; Sensitivity analysis|
|How can we improve methods to obtain relevant soil data?|
|4||Can we measure soil properties more efficiently?||Soil sensing ; Pedotransfer functions ; Translation of qualitative soil information ; Participatory approaches and citizen science ; Sampling design ; Measurement error ; Multi-source data integration|
|5||Can we develop workable techniques to derive predictions of soil characteristics at scales appropriate for modelling and decision making, by up- and downscaling observations in 3D space and time?||Upscaling and downscaling ; Sampling support ; Change of support ; Temporal scale issues in modelling change ; Validation for change of support|
|6||Can we incorporate mechanistic pedological knowledge in digital soil mapping?||Pedological knowledge ; Extrapolation ; Qualitative soil information ; Mechanistic modelling ; State-space modelling ; Uncertainty in mechanistic knowledge|
|How can we improve our ability to address demands by soil users?|
|7||How to recognize, quantify and map soil functionality?||Soil function and services ; Citizen-observation of soil functions ; Land evaluation ; Multivariate mapping ; Bio-physical models ; Co-building of functions with end-users|
|8||Can we find ways to connect pedodiversity to soil biodiversity, and translate the connections to relevant soil services and soil management practices.||Pedodiversity ; Pattern of soil biodiversity ; Scaling issues in pedodiversity ; Taxonomic distance ; Hyper-variate data of soil biodiversity ; Sensing for microscale biodiversity|
|9||Can we find ways to express the uncertainty of predictions of soil properties or class allocations which are meaningful to the users of those predictions?||Uncertainty quantification; Value of information ; Risk assessment ; Uncertainty and decision-making process ; Communication of uncertainty ; Decision theory and support scale|
|10||How to quantify soil contributions to ecosystem services with a framework enabling both local and regional soil management?||Ecosystem services ; Local and regional soil management ; Empirical land evaluation schemes ; Soil health and security quantification ; Soil contributions to realizing the SDG|
Deadline October 15th 2023
Please limit abstracts to 350 words
- Abstract Submission Deadline: October 15th, 2023
- Notification of Acceptance: November 15th, 2023
- Early Bird Registration Deadline: December 15th, 2023
- Deep learning for soil spectroscopy
Organizer: José Padarian, Wartini Ng. The University of Sydney.
Potential duration: 6 hours
Maximum number of participants: 30
Participant requirements: Laptop, basic programming skills. Experience using
Topic: This workshop will explore the use of deep learning for soil spectroscopy, in
particular the use of convolutional neural networks. Neural networks offer a very
flexible framework for modelling that has shown great potential for soil predictive
modelling. In particular, we will explore: 1) The basic concepts behind neural networks. 2) How to pick a neural network architecture, layer types and loss functions to tackle different problems including regression, classification, and multi-task regression (predict multiple properties simultaneously). 3) How to train a neural network to avoid overfitting. 4) How to assess the uncertainty of deep learning models.
- Assessment of spatial patterns of soil properties predictions
Organizer: David Rossiter (ISRIC)
Potential duration: 4 hours
Maximum number of participants: 24 Participant requirements: Laptop
Topic: We present methods to evaluate the spatial patterns of the geographic distribution of soil properties as shown in gridded maps produced by digital soil mapping (DSM) at global, national, and regional scales. We compare them to spatial patterns known from detailed field surveys or known to local experts but not represented (yet) on maps. The methods will assess whole-map statistics, visually identifiable landscape features, level of detail, range and strength of spatial autocorrelation, landscape metrics (Shannon diversity and evenness, shape, aggregation, mean fractal dimension, and co-occurrence vectors), and spatial patterns of property maps classified by histogram equalization. Histograms and variogram analysis revealed the smoothing effect of predictive models. The workshop will provide a tutorial with examples for the USA but the participants will be able to use the provided examples to test on other areas.
- Containers for reproducible Digital Soil Mapping at different scales
Organizer: Laura Poggio (ISRIC), Giulio Genova (ISRIC)
Potential duration: 4 hours
Maximum number of participants: 24
Participant requirements: Laptop
Topic: In this workshop we will introduce the topic of computational containers for Digital Soil Mapping (DSM) and describe experiences on different infrastructures (HPC, personal laptop, commercial clouds). We will explain the main pros and cons of using containers for DSM projects and different needs for computational resources. We will provide a short tutorial for the participants to get started using containers technologies on their own laptop.
- Algorithms for Quantitative Pedology
Organizer: Dylan Beaudette (USDA-NRCS)
Potential duration: 4 hours
Maximum number of participants: 20
Participant requirements: personal computer, R (version 4.2.2 or greater), pre-installation of R packages (sent out ahead of time).
Topic: This four-hour workshop will cover several R packages of interest to scientists and engineers who work with soil morphology and laboratory characterization data on a regular basis. These packages are maintained by USDA-NRCS SPSD staff and serve as key infrastructure for the routine work of the U.S. soil survey program. Participants will learn how to create and manipulate SoilProfileCollection objects from their own data or U.S. Soil Survey data, generate soil profile sketches, aggregate over groups, and visualize many different aspects of soil profile data. There will also be several demonstrations of the Numerical Comparison of Soil Profiles (NCSP) algorithm for numerical classification of soil profile data. Case studies and example data will be based on public data sources made simple using the soilDB package. Examples include: Kellogg Soil Survey Laboratory (point data), Soil Data Access (detailed soil survey data), Web Coverage Services supporting thematic soil property and condition mapping, as well as the Soil Climate Analysis Network (SCAN). The first half of the workshop will focus on a tour of the aqp, soilDB, and sharpshootR packages with numerous case studies that demonstrate a range of simple to advanced analyses. The second half of the workshop will include time for question/answer session, discussion of package and algorithm design choices, feature requests, and 1:1 assistance with users’ specific data. A primer on these packages along with a simple pre-course assignment will be sent out one month before the workshop. This will ensure that all participants will have a fully functional R environment with the latest versions of required packages. All code, documentation, and follow-up resources will be posted to a GitHub repository for reference after the workshop and conference.
The conference will be located on the campus of New Mexico State University at the Corbett Center Student Union. The following map shows the location of the Corbett Center Student Union. The meetings will be held on the second and third floors. This location is within walking distance of several restaurants.
The weather in Las Cruces in early February is usually mild. Daily temperatures range between overnight lows of 30F (-1 C) and daytime highs of 70F (21 C). Expect bright and sunny conditions as the chance of precipitation is expected to be near 0% (actually, the chance of precipitation is almost always near 0% in Las Cruces!). However, Las Cruces can experience windy conditions in the spring. Windy conditions generally do not begin until mid-March, but windy conditions are possible.
A 10-day weather forecast for Las Cruces can be found at https://www.wunderground.com/forecast/us/nm/las-cruces/KLRU
The nearest airport is the El Paso, Texas international airport. It is an approximately 1-hour drive from the El Paso airport to Las Cruces. The easiest way to reach Las Cruces from the El Paso Airport is to use the Las Cruces Shuttle service (https://www.lascrucesshuttle.com/shuttle). Please contact the Las Cruces shuttle service to book a ride.
It may also be possible to use Lyft or Uber to schedule rides.
The following hotels are all within walking distance of the conference venue. Please be aware that Las Cruces has very little public transportation if you choose a different option.
- Courtyard by Marriot
- Holiday Inn Express
- Ramada Palms
- Quality Inn and Suites
- Hilton Garden Inn
- Comfort Suites
- Sleep Inn