 |
|||||
 |
 |
 |
 |
 |
 |
Making carbon crediting really work for farmers, Part 3
Making in-field carbon measurement a reality.By Elaine Viglione In the last article, we discussed how a Mobile Field Laboratory that measures soil carbon with hand-held sensors can minimize the cost and time involved with traditional laboratory soil carbon analysis. Since that time, we have generated our first sensor-derived field carbon estimates for the Farming Systems Trial and mapped them instantaneously, without ever leaving the field. With this development, we come one step closer to making in-the-field carbon measurement a reality and have identified some of the next challenges in research and implementation of this system. Why do we care about measuring soil carbon? Because carbon crediting programs are based on soil’s beneficial sequestration potential; i.e. the more carbon we can hold in the soil, the less there will be in the atmosphere. Another reason is that the presence of carbon is a key indicator of healthy soil. Having accurate and plentiful measurements of soil carbon allow us to better understand these relationships, and also support two stages of a carbon crediting scheme. The first stage is the underlying research, through which field trials and scientific results are used to guide policy-making that rewards farming practices that accumulate soil carbon. When soil carbon data is more reliable and accessible, we can better study and understand the true effects of various practices. The other stage is important after policies are made, and involves monitoring and verification of soil carbon accumulation. There are different schools of thought as to whether carbon-building practices should be verified via a) documenting that the practice is taking place and attaching a set carbon increase to that practice, or b) getting on the ground and measuring the soil for actual carbon change before and after a practice is implemented. There are proposals involving both types of verification out there in the potential carbon crediting community. If a verification program requires soil carbon measurement, this will demand a system that can gather a large number of samples quickly and efficiently, conduct soil carbon assessments, characterize their accuracy, and adapt to highly variable landscapes. With support from the Pennsylvania Department of Environmental Protection’s Pennsylvania Energy Development Authority (DEP-PEDA) and in collaboration with the Penn State University Department of Crop and Soil Sciences, Rodale Institute started laying the groundwork for the Mobile Field Lab in the nearly 30-year-old Farming Systems Trial (FST) this spring. This article will take you through many aspects of the project “Rapid, Cost-effective Soil Measurements for Accurate Agricultural Carbon Crediting” including sampling infrastructure upgrades, analytical enhancements, preliminary validations of the handheld sensors, system integration, and new challenges. Upgrades to the Flagship Farming Systems TrialThe Farming Systems Trial (FST) has been studied for nearly 30 years, and to better support continuation of this work, the first technology upgrade to our Mobile Field Lab included two enhancements that allow better long-term sampling and analysis of carbon in FST and on farms. First, the installation of a field sampling grid, comprised of underground markers that are buried below the plow line, will allow us to precisely revisit sampling locations in the trial at 1, 5, even 20 years. We will use this prototype to create reference surveys for farms that we may not visit as often, because the markers, which will stay in place regardless of routine farm operations, will allow us to easily conduct accurate follow-up surveys.
This field-based marker grid will be coupled with new high-resolution Global Positioning (GPS) equipment, to assist in surveying the marker locations in FST, on farms, and to fill in between the markers with finer carbon sampling grids. More importantly, the GPS allows us to compile time-date-location logs of each carbon sampling location, which is essential to the instant mapping capability of the Mobile Field Lab. Enhanced Analysis ToolsEnhanced analysis tools, such as a Geographic Information System (GIS), have been added for several functions: 1) to improve our understanding of the FST landscape, 2) to prototype the tools that allow us to collect on-site spatial information to use in adapting sensor calibrations to on-site conditions, and 3) to convert field-collected measurements to real-time carbon maps and carbon stock assessments. Because most laboratories, as well as the handheld sensors in the new Mobile Field Lab, measure soil carbon in terms of concentration, we need to understand key physical soil properties in order to transform these concentration readings into a reliable estimate of the total carbon mass present in the soil. So our first detailed GIS study is development of a map that characterizes the spatial composition of FST, including spatially-intense sampling and mapping of bulk densities and rock fragment concentrations.
Figure 2 above provides a snapshot of one of these properties, the disposition of rock fragments in the Farming Systems Trial. Statistical analysis (using the GeoR package of the R Project software) allowed us to model FST’s rock fragment concentration, which was collected over a limited number of points, as a full surface that estimates fragment concentration at any point in FST. Handheld Sensors and PrinciplesSince high-carbon soils are generally known to be darker than low-carbon soils, the Mobile Field Laboratory uses handheld spectrometers that estimate carbon concentration by measuring the visible, near-infrared, and mid-infrared portions of the electromagnetic spectrum. Early demonstrations of this equipment at Rodale Institute workshops in May and June illustrated some of the principles that can be exploited when using the optical spectrometer to estimate soil carbon concentration, as are shown below. We plotted some extreme soil examples to illustrate how the spectrometer functions. Figure 3 shows, in yellow, the visible spectrum output from conventionally-farmed FST soils, which our historical data show have evolved to be the lowest carbon group in FST during the past 30 years.
The brown and black curves are from garden soil and compost respectively, which have much higher carbon concentrations. One workshop participant brought in their own compost sample to be measured with the handheld spectrometer (blue curve). The proximity of the home owner’s compost curve to the compost reference curve, particularly in the higher wavelengths, suggests that these curves might be used as signatures to evaluate soil properties like carbon. Modeling and ValidatingTaking this idea to the next level, we developed a processing and sampling protocol to generate carbon concentration models based on soil samples archived from FST over the last 30 years, samples that had already been analyzed for carbon through traditional laboratory processes. Figure 4 illustrates the incremental relationship between carbon concentrations versus response in the visible spectrum.
There are many ways to transform these patterns into soil carbon estimates, and so the next step was to set up validation testing at Rodale Institute. In this process, we compared various calibration models to evaluate their performance in predicting carbon concentration, relative to one another. Team members from Rodale Institute and Penn State contributed calibration models to this validation framework. In addition to varying the models, we also varied the wavelengths of the spectrum used for carbon concentration estimation. We varied the wavelengths in two ways: both randomly and selectively. The latter means that we used prior knowledge about which portions of the spectrum are not relevant to carbon concentration. A simple example is a model which leaves out the ultraviolet wavelengths. Figure 3 is a useful illustration of the impact of outdoor measurement on the ultraviolet wavelengths in these readings. The reference curve for compost (black curve) was developed indoors, but the workshop participant’s compost (blue curve) was measured outdoors. As such, the high readings in the lowest end of the spectrum of the workshop participant’s curve shows the effect of outdoor sampling on the ultraviolet feedback. This validation showed that some models are more promising than others, and certainly some parts of the spectrum are more useful than others. A new round of this validation is now under way at Rodale and includes: larger sampling of the archived FST soils, comparing archived samples with actual field samples, using multiple sensors (e.g. visible and near infrared estimating together) and incorporating landscape variations. Again both teams have contributed calibration candidates to these validations. This new round of validation will yield carbon prediction performance metrics so that we can, in turn, improve the Mobile Field Laboratory’s ability to report accuracy in the field. Taking it to the FieldSo far we have described validating the spectrometers with data from the Rodale site only, using both fresh and archived soil samples. However, the Mobile Field Lab needs to be adaptable to a variety of field conditions, and more importantly, to new sampling locations at different farms. There are many soil properties, in addition to carbon, that affect the output of the spectrometers. Because the instruments are measuring in the visible spectrum, other properties of the soil can influence either the sensor reading, the carbon level in the soil, or both. For this reason, a site with a completely different soil type could be better suited to a different model, or at least warrant a re-fitting of a base model. For example, we have found that one of the experimental blocks in FST which is higher in iron content shows a different signature in the visible spectrum than other experimental blocks, even when the samples have similar carbon levels. The unique reaction prompted by certain soils means we need to generalize our approach to a multitude of sampling locations and farms. As such, an essential part of the next level of validations at Rodale Institute includes a rapid, on-site, assessment feature that will be included for each new site participating in a carbon measurement program. A critical part of a site’s pre-assessment is determining how many traditional laboratory samples we will need to select the best carbon model for new locations from the calibration candidates we have in our tool bag. The idea is to use a minimal number of traditional lab-derived carbon measurements to support the cheaper, more plentiful measurements we can obtain from the spectral Mobile Field Lab. We are incorporating data from Penn State and other locations to refine model selection into a dynamic process embedded in the Mobile Field Lab. Putting it all TogetherWe have described data collection in the field using a number of devices (GPS, sensors, GIS, markers, etc.). How long will it take to put all of this information together into a single summary format that’s useful for mapping and analysis? Not long, thanks to the Penn State team’s development of a universal barcode scheme to manage the data. All of the components described above are fed into a central field laptop, using Bluetooth or USB connections in the field. Penn State has pioneered a set of field computing, data processing and internal referencing forms that allows us to link all of the information via a barcode. This barcode is also linked to the physical soil samples for field and laboratory analysis later. Rodale is constructing a central database that pushes/pulls this information to and from the statistical and GIS software and generates the site assessment/model selection, reports prediction accuracies, and creates field maps and carbon stock assessments. In preparation for Rodale Institute’s July 16th Field Day, we demonstrated a prototype that took a soil core from the ground, obtained a visible sensor reading, fed it into a model that generated a carbon concentration estimate, and plotted it accurately onto a field map, all working right in the field. It also generated a carbon stock estimate (in metric tons per hectare) using previously collected spatial soil composition data. Looking ForwardFor the next report, we look forward to sharing additional results on the spatial composition sampling, the Mobile Field Laboratory model validations, the Mobile Field Lab’s performance in predicting soil carbon in the field, and on a protocol for on-site validation and adjustment when taking this system to a new site. Stay tuned! Elaine Viglione is Analysis and Mapping Coordinator for Rodale Institute Part I, Part II, Part IVThis research was supported with funding by the Pennsylvania Department of Environmental Protection’s Pennsylvania Energy Development Authority (DEP-PEDA), grant number 41000455440, “Rapid, Cost-effective Soil Measurements for Accurate Agricultural Carbon Crediting.” |
