May 3, 2017

SCILSS – Soil Carbon Integrated Landscape Sampling System

Led by PhD student Dan Kane, a group of researchers at Yale F&ES has developed the Soil Carbon Integrated Landscape Sampling System (SCILSS) – a low-cost protocol for rapidly measuring soil carbon across large landscapes at fine spatial resolutions. The inexpensive nature of this methodology lends land managers the ability to look at impacts of management decisions on below-ground carbon at broad extents and frequent time intervals. For these reasons we believe that SCILSS has the ability to change the way we understand and manage for carbon in rangeland systems.

Pilot Projects:
Tensleep Preserve, WY | Hart Mountain, WY | Red Canyon Ranch, WY | Ucross Ranch, WY | TwoShoes Ranch, CO | Michigan Farms | Wolfsbane, ME | Tomkat Ranch, CA

 
 
Summary of SCILSS Methodology

Soil is the largest terrestrial carbon store, holding roughly 3000 Gt of carbon, or more than 80% of terrestrial carbon. Estimates suggest that human disturbance has decreased this reservoir by 55-78 Gt of carbon, and that soils are likely to lose still more as climate change and land conversion accelerates decomposition processes. Recapturing lost soil carbon could be an important strategy in mitigating climate change. In a comprehensive review of regional case studies, researchers found that increasing surface soil carbon stocks by just 0.4% globally could offset 20-35% of global greenhouse gas emissions. Research has identified land management practices that increase carbon transfer to soils by increasing plant growth while also minimizing losses of soil carbon by reducing disturbance. Non-profits and government agencies have begun to look for ways to incentivize managers to transition to these management practices, but there are two major impediments to their broad-scale adoption:

1.) Uncertainty as to whether landscape management changes increase soil carbon. Some practices increase soil carbon in some areas but not others. Identifying the range of site characteristics over which management increases soil carbon is necessary for guiding adoption of specific practices tailored to particular landscapes.
2.) Can an increase be measured at the scale of landscape management? Assessing soil carbon stocks currently requires gas chromatography and elemental analysis. These laboratory-based methods are highly accurate, but also time consuming and expensive. High costs limit the number of samples managers can take, meaning quantification happens infrequently and is only done on a small portion of a farm or ranch. The result is that current soil inventory methods lack the spatial and temporal resolution needed to accurately quantify soil carbon stocks across large scales to adequately detect change over time and understand the impacts of management.

We have recently developed a soil carbon measurement protocol that makes use of low-cost field spectrometers. These affordable, pocket-sized devices measure soil carbon using the reflectance of soils in the visible and infrared spectra. As carbon content increases, a soil’s color darkens, giving it a slightly different spectral signature than soil with lower carbon content. Standard benchtop spectrophotometers used in similar work cost $3000-$10,000 and are not portable, whereas this device can be produced for less than $500. Further, remote-sensing data that correlates well with background concentrations of soil carbon can be integrated with data from this device to improve the accuracy of the statistical models used for measurement and produce high-resolution, localized maps of soil carbon stocks. This device and protocol could address both of the major impediments to adoption outlined above.

At a given site, we first establish a baseline by sampling soil throughout the site to capture a range of soil carbon contents using traditional, highly accurate lab-based techniques. Those data are then used to build statistical models relating lab-measured soil carbon levels to the data collected with the field spectrometer. After the initial site-specific calibration, carbon content can be determined in the field using only the pocket spectrometer, dramatically reducing sample collection time and cost and allowing managers to sample more frequently in the future and over broader areas. With these reductions in cost, managers can quantify soil carbon at the landscape scale at more frequent intervals, removing the concerns about the current inability to monitor soil carbon responses as new managements are adopted.

Our initial field tests are promising. Working with a first-generation field spectrometer we can predict soil carbon content within 1% accuracy. Extracting data from existing geographic datasets, such as vegetative productivity and from SoilGrids, we were able to improve accuracy to within 0.88%. In comparison, past researchers using similar techniques on more expensive laboratory-based instruments have achieved accuracy of 1-3%. Our models could also be improved by integrating spatial data from remote sensing satellites or newly available high resolution aerial photography (i.e. drone flyovers). With sufficient accuracy we could also interpolate across a given landscape to produce hyper-local maps of soil carbon stocks on a given landscape.

Furthermore, the spectrometer is built on an open-source hardware/software platform and is capable of integrating via Bluetooth with sample collection software that runs on the Android platform thus automatically providing sample coordinates. Given these features we believe we can develop this device and protocol into a methodology that is low cost and that land managers could easily use. Reduced sampling time and cost would allow managers and researchers alike to take more samples over more area, providing real time monitoring of changes to soil carbon stocks and high-resolution maps of soil carbon at the scale of farms or ranches. That level of monitoring capability would allow us to understand what management practices are effective and where, providing the foundation for soil carbon offset markets and effective extension programming.