Ecological Monitoring Scientific Research Round-Up
Digging Deep with #DeSci
Since its inception, Regen Network Development Inc. has been committed to incentivizing a quality uplift of the Monitoring, Report and Verification (MRV) processes involved in ecological outcomes reporting and carbon accounting, relying on scientific research and development (R&D). The internal science team at RND regularly carries out research activities targeted to create open-source community tools and resources that can address key research questions associated with the MRV processes.
In 2021, we participated in a Working Group led by Repliculture LCC which received grant support from the Ecosystem Services Marketing Consortium (ESMC) to co-author several scientific research reports where we address the potential of satellite remote sensing, proximal sensing, direct measurement, and modeling tools and technologies to improve in-field estimates of soil organic carbon (SOC) stock changes. You can find the reports here!
The ESMC Scientific Research Reports at a Glance
Soil carbon stocks quantification has historically been implemented using direct measurement techniques (i.e. extracting soil cores from a field and analyzing soil organic carbon (SOC) and bulk density (BD) in a laboratory setting). These methods are still widely used and are largely seen as the standard for comparing alternative methods. In the last ten years, new methods to estimate SOC have evolved to include alternative methods of in-lab measurement, in-field measurement, and on-the-go measurement, the latter two categories being referred to as proximal soil sensing (PSS). These alternative PSS methods offer the possibility of more cost-and time-effective measurement of soil carbon and soil carbon stocks, but they generally do not possess the same depth of historical research to support their methods.
In this report, we review direct measurement and proximal sensing methods and their potential costs, accuracies, and commercial availability to date. We describe common methods of measuring SOC and BD in the lab and then compare those with lab-based methods that use visible Near-infrared (Vis-NIR) and mid-infrared (MIR) reflection spectroscopy to estimate SOC. Finally, we compare these approaches through the lens of utility for SOC stock change quantification over the immediate and longer-term time frames, with a focus on the field scale.
The profusion of geospatial data layers over the last 20 years has provided a wealth of information that can be potentially used to predict SOC concentrations and bulk density (BD) , the two primary inputs for calculating SOC stocks. Simultaneous increases in computing power have enabled these data layers to be statistically combined using simple and complex algorithms to improve predictions. Growth in academic, peer-reviewed modeling literature has mirrored these increased capabilities, resulting in a huge array of layer combinations and statistical methods, each with varying accuracies across agricultural systems.
Despite this progress, spatial modeling of SOC stocks is still constrained by availability of ground-truth data, which are needed to calibrate and validate each model’s efficacy. Therefore, while these approaches are being incrementally improved and tested, there will still be a need for fieldwork to collect concrete, trusted measurements directly from farms and ranches in order to generate high-quality credits.
In this report, we review academic modeling literature, in which we assess geospatial data layers and algorithms used for prediction. Next, we evaluate the efficiency of different stratification approaches, most of which use geospatial data layers as input. Finally, we provide recommendations on the use of modeling algorithms and stratification sampling techniques as a way to estimate SOC stocks at the field scale.
Figs. 1 and 2 summarize those covariates that resulted in the most relevant (i.e., with high explanatory power ) for predicting SOC at the field scale. Elevation, slope, NDVI, and the topographic wetness index (TWI) all tended to perform well. The algorithms that were more successful in predicting SOC at the field scale based on covariates were Multiple Linear Regression (MLR), Ordinary Kriging (OK), Random Forest (RF), and Artificial Neural Network (ANNS)
To enable payments for sequestered carbon, scalable technologies must be put in place that accurately and cost-effectively quantify changes in SOC stocks. Remote sensing (RS) is a promising tool for measuring SOC stocks at the field or parcel scale. In fact, we explored this when developing our CarbonPlus Methodology for GHG and Co-Benefits in Grazing Systems
In this report, we assess the accuracy and efficacy of multiple remote sensing technologies for measuring SOC. We reviewed more than 100 papers and technologies. We also briefly outline the challenges inherent to satellite-based RS platforms, including low signal-to-noise ratios, atmospheric distortion, geometric distortion, bi-directional reflectance, cloud cover, and obtaining bare-earth imagery (most of the evaluated studies selected RS imagery during periods when there was a high bare-earth cover). We conclude that any RS-based solution will need to minimize and/or correct for these issues to provide accurate assessments of SOC.
Based on this meta analysis, there’s potential to achieve good accuracy for the estimation of SOC stocks when using satellite remote sensing. Furthermore, to advance the ability for RS technologies to be used for reliable SOC quantification, we recommend development of a systematic, statistically robust sampling campaign that is implemented in a variety of bioclimatic contexts.
In the three interim reports preceding this final report, we reviewed in detail tools and technologies designed to improve in-field estimates of soil organic carbon (SOC) stock changes. From broad-scale approaches based on remote sensing to fine-scale sampling methods using penetrometers, all possible options were considered for improving accuracy and reducing costs. Each possible technology comes with its own unique trade-offs; there are no comprehensive solutions to quantification that eliminate the need to balance benefits and costs.
To help guide ESMC’s approach to navigating this challenging landscape, this final report provides a summary of our previous findings and an integrated analysis of all the trade-offs associated with estimating SOC stock changes. Beginning with a review of the inputs to the stock equation and sources of uncertainty, we describe the categories of tools available to estimate SOC concentrations, bulk density (BD), depth of sampling (d), and rock fragments (RF). Next, we perform a Monte-Carlo analysis of the impact of accumulated uncertainties on the probability of detecting stock changes and weigh those probabilities explicitly against costs. Finally, we synthesize the analysis with general observations on the most promising tools and technologies that may be available in the near future.
#DeSci at Regen Network
Open science research is at the heart of Regen Network’s role furthering ecological regeneration. We believe that public research furthers our ability to support communities of practice, especially the 40+ methodology developers partnering with Regen Registry, our web3 native ecological asset origination platform. Our #DeSci team can be reached for research, development, and scientific partnerships via firstname.lastname@example.org.
About the Author:
Dr. Gisel Booman is an experienced researcher who has been leading the Science Team at RND Inc. since 2018. Before co-creating with Regen Network, she received her degree in Biological Sciences in 2007 from the National University of Mar del Plata (UNMdP), Argentina. Through a fellowship from CONICET she trained for her PhD in Landscape Ecology. From 2012 to 2018, Dr. Booman led GIS and remote sensing projects, including the development of an index for the identification of priorital grasslands to be preserved in South America (Alianza del Pastizal / IADB) and the identification and delineation of wetlands in the Rio Leon catchment, Colombia. In 2013, Gisel became an adjunct university professor, teaching GIS to agronomists at the UNMdP. Between 2015 and 2017, she was awarded with a postdoc fellowship by CONICET, to contribute to the UE project iMetLand. She has authored several scientific publications and is the author of the CarbonPlus Grasslands Methodology, published on Regen Registry, with credits sold to the Microsoft Moonshot. Dr. Gisel Booman can be found on Twitter @BoomanGisel or LinkedIn.