Cooperative Institute for Research in the Atmosphere

We propose to use modern remote sensing data products from NASA and ESA to model regional carbon dynamics for agricultural lands, with the goal of improving the national GHG inventory for the United States. U.S. agricultural lands have been a modest sink for CO2 during the past few decades, and there is a re-emerging interest in encouraging C sequestration in agricultural lands in national policies as well as carbon markets such as Nori and Indigo Ag. We will build on existing research applications of solar induced fluorescence (SIF) for estimating crop production to improve estimation of C input to soils, which is a major driver of soil C changes modulated by seasonal and interannual variations in weather and climate. Furthermore, C inputs to soil reflect the impacts of practices such as selection of higher producing crop varieties, cover crops, improved nutrient management, irrigation management, and other practices. In the process, we will extend the application of SIF to crucially under-studied semi-arid regions where SIF can provide a unique constraint on model-based analyses of crop production, C input to soils, and interannual variability in regional CO2 fluxes between agroecosystems and the atmosphere.
Over the last decade, SIF has emerged as a powerful predictor of plant photosynthesis. Mechanistically, SIF has direct connections to photosynthesis at the leaf level in contrast to measurements like NDVI and EVI, which have additional issues of reaching saturation. It has been shown to be a more powerful predictor of crop yield as well as net (NPP) and gross primary productivity (GPP). Moreover, it is clear from the governing equations for SIF that this relationship should deviate from linearity under conditions of high water and/or temperature stress. For example, crops in rainfed and limited irrigation systems are often influenced by short term stress periods that are not always apparent with traditional vegetation indices. The accumulated effect of such stress can have a significant influence on production and yield on the annual level, and in turn, influence C inputs to soils and regional CO2 fluxes.
In collaboration with USDA-ARS scientists, we propose to investigate the relationship of SIF to productivity and yield in crops under stressed conditions. We will instrument two established USDA field operations, the Limited Irrigation Research Farm (LIRF) in Greeley, CO and Central Great Plains Research Station in Akron, CO, with tower-based SIF instruments (PhotoSpec). This will complement a suite of ongoing measurements, including drip water irrigation, meteorology, sap flow, CO2 eddy covariance, in addition to vegetation indices and routine spectral imagery from once a day autonomous drone flights. We will further test leaf level SIF modeling from the SiB4 model to better understand the relationships between SIF measurements and photosynthesis estimates from sap flow and eddy covariance, biomass, crop stage, soil moisture and temperature. Along with previous algorithm development for the mesic croplands such as the corn-soybean production in the Midwest, we will develop a SIF-based production algorithm within the DayCent Ecosystem Model to extrapolate the local scale modeling results to croplands in the western United States. We will use SIF from OCO2/3 and TROPOMI, as well as targeted OCO-3 flyovers to produce a new SIF-based product of regional CO2 fluxes in agricultural lands to enhance our understanding of interannual variability on flux patterns. We will benchmark improvements based on comparisons to the current approach in the U.S. GHG and anticipate that the SIF-based method will be used in future US. GHG inventories. The driving principles and philosophy combined with the global coverage of the aforementioned satellites could have broad applicability across water-limited agricultural regions around the world, e.g., wheat in Russia, China, and India, maize in China, Brazil and India.

We propose to use a combination of bottom-up and top-down methods to constrain the carbon budget of Brazil. We will build an inversion system, capable of resolving the spatially heterogeneous fluxes and complex patterns of atmospheric transport for Brazil, using the OLAM general circulation model (GCM) with a “refined mesh” capability, in combination with an improved process-based CO2 flux prior. Bottom-up flux priors from the SiB4 model will be improved to reflect better understanding of evergreen forest seasonality based on results from the OCO-2 Flux MIP inversion results (Peiro et al 2022, Byrne et al 2022). More realistic simulation of multiple rotation tropical crops will be added as well. Global CO2 flux outside of South America will be constrained by the OCO-2 Flux MIP ensemble. A synthesis Bayesian approach will be used to optimize CO2 fluxes and inferences made on the ‘state’ geopolitical level of Brazil, complete with uncertainty.
The work we propose targets the CMS Flux prototyping activity. By effectively producing a downscaled regional flux estimation system complete with uncertainty, we are proposing product improvements, refined characterization and quantification of errors and uncertainties as it relates to the CMS Flux prototyping activity. At the same time, we are advancing a product which can be used to validate/evaluate portions of the national emission estimates for Brazil as they relate to the Global Stocktake. Furthermore, the proposed research addresses several previously identified challenges for the CMS flux community, including lack of atmospheric observational data, need for regional inversions with robust specification of lateral boundaries, clear uncertainty estimates, and more highly-resolved transport.
Bottom-up flux estimates will be improved through collaboration with Brazilian colleagues (crops) and leveraging separately-funded improvements in the representation of tropical forest. The improved prior fluxes will be transported using the high-resolution OLAM simulations leveraging the collaboration of local Brazilian meteorologists. Final optimized fluxes will be evaluated against withheld satellite XCO2 data, aircraft profiles sampled by collaborators in Brazil and at NOAA GMD, as well as existing CMS-Flux system estimates. Looking into the future, this framework, established and tested in a well-observed country like Brazil (for the tropics), can be applied in more sparsely sampled or politically volatile regions around the world.

High-quality observations of XCO2, the total column CO2 dry air mole fraction, from space-based and ground-based remote sensing have provided key insights into carbon cycling globally, although questions of local to global significance remain unanswered. More robust carbon flux estimates may benefit from recent advances in the spatiotemporal timescales over which CO2 can be inferred via remote sensing. First, the timescales over which we can observe XCO2 have broadened from diurnal (OCO-3) to decadal (OCO-2 and GOSAT). Second, algorithms have been developed to infer two pieces of vertical information from near infrared spectra so that an upper and lower tropospheric partial column can be inferred (Kulawik et al., 2017; Parker et al., 2023), and GOSAT v3 retrievals now include upper and lower tropospheric products. Most recently, the Temporal Atmospheric Retrieval Determining Information from Secondary Scaling (TARDISS) algorithm (Parker et al., 2023) separates ground-based Total Carbon Column Observing Network (TCCON) data into a lower (0-2km; LT) and upper (>2km; UT) partial column. Detection of diurnal variations and retrieval of partial columns may help isolate local-scale (e.g., regional fluxes) versus large-scale (e.g., transported) contributions to column CO2, since the LT partial column is more directly connected to surface exchange and diurnal timescales may be sufficiently short such that atmospheric transport impact is minimized. This proposal builds on previous research by the PI to understand the information contained in variations in column CO2. Here, we turn our attention to partial column data, both in terms of the sensitivity to local fluxes and in terms of constraining atmospheric transport, which remains a critical source of uncertainty in flux inference. Our research is centered around careful analysis of novel TARDISS retrievals, then channeling that assessment into an analysis of a hierarchical Bayesian inversion framework (WOMBAT; Zammit-Mangion et al., 2022; Bertolacci et al., 2022) that constrains fluxes based on OCO-2 and OCO-3 XCO2 observations.

Unique to this proposal is that we will conduct WOMBAT inversions using two different atmospheric transport models, allowing us to evaluate the impact of transport in the context of TARDISS partial columns. Because atmospheric transport error is difficult to quantify, our use of two models with documented differences (Schuh et al., 2019, 2023) within the same inversion framework provides a unique method to model that uncertainty. Our WOMBAT inversions will participate in the Flux MIP (following Crowell, et al., 2019, Peiro et al., 2022, Byrne et al, 2023). From this broader set of models, we will develop emergent constraints on posterior fluxes derived from the MIP spread in simulated partial columns and optimized fluxes, providing improved uncertainty quantification. Our research will accomplish carbon cycle science through enhanced use of TCCON validation data in conjunction with space-based observations from the OCO missions. We anticipate that our inversions will be an important, and unique, contribution to the Science Team, since WOMBAT is among the few inverse frameworks allowing readily available posterior uncertainty on CO2 flux estimates.

NASA invests hundreds of millions of dollars every year on remote sensing platforms that aim to quantify the carbon cycle, whether through measurements of atmospheric composition or the land surface. Composition measurements – such as CO₂ and CH₄ – cannot be used directly but must be interpreted through inverse models of different complexity to estimate surface sinks and emissions, from global to urban scales. At present there are few venues where future scientists can get trained in atmospheric inverse modeling, since it involves an intersection of various disciplines including (but not limited to) applied mathematics, software engineering, biogeochemical cycles, geostatistics and isotope geochemistry. Most of today’s inverse modelers came from one of those fields and were fortunate to work under a mentor who filled in their gaps in knowledge. This unfortunately is a bottleneck in developing a future workforce of skilled inverse modelers who can interpret multiple data sources to construct surface flux estimates, both for understanding the carbon cycle and shaping policy. For researchers in the field, this shortage shows up as the inability to find young scientists to take on the challenge of developing new techniques to assimilate the various current and future satellite data sources of greenhouse gases (GHGs).
We propose to organize an annual summer school in atmospheric inverse modeling for graduate students to train young scientists to interpret and use atmospheric GHG data. There is ample precedent of such summer courses in other fields which require cross-disciplinary knowledge, such as the FLUXNET summer school in Niwot Ridge, CO on using flux tower data and two summer courses in Salt Lake, UT on isotope geochemistry . Similar short courses are routinely held in Europe, such as in Fréjus, France in 2022 (inverse techniques) and Grenoble, France in 2023 (atmospheric modeling). We propose to fill a gap that currently exists for atmospheric inverse modeling in the US. We plan to host the first two summer schools in Ft. Collins, CO.
A summer school in atmospheric inverse modeling will also have an important co-benefit. The mathematical methods involved are very similar to those required for retrieving GHG concentrations from space, another area in need of workforce development. Despite the multi-billion-dollar investment worldwide in launching GHG-sensing satellites, there are only a handful of groups worldwide (and only two or three in the US) that possess the expertise and resources to estimate CO₂ and CH₄ concentrations from the spectra that these satellites measure. This limits the number of future satellite retrieval scientists to those who have spent considerable time in one of those groups. An inverse modeling summer school also has the potential to address that workforce shortage by giving graduate students the fundamentals necessary to become GHG satellite retrieval scientists.