A data assimilation framework for constraining upscaled cropland carbon flux seasonality and biometry with MODIS

Abstract

Agroecosystem models are strongly dependent on information on land management patterns for regional applications. Land management practices play a major role in determining global yield variability, and add an anthropogenic signal to the observed seasonality of atmospheric CO₂ concentrations. However, there is still little knowledge on spatial and temporal variability of important farmland activities such as crop sowing dates, and thus these remain rather crudely approximated within carbon cycle studies. In this study, we present a framework allowing for spatio-temporally resolved simulation of cropland carbon fluxes under observational constraints on land management and canopy greenness. We apply data assimilation methodology in order to explicitly account for information on sowing dates and model leaf area index. MODIS 250 m vegetation index data were assimilated both in batch-calibration for sowing date estimation and sequentially for improved model state estimation, using the ensemble Kalman filter (EnKF), into a crop carbon mass balance model (SPAc). In doing so, we are able to quantify the multiannual (2000–2006) regional carbon flux and biometry seasonality of maize–soybean crop rotations surrounding the Bondville Ameriflux eddy covariance site, averaged over 104 pixel locations within the wider area. (1) Validation at the Bondville site shows that growing season C cycling is simulated accurately with MODIS-derived sowing dates, and we expect that this framework allows for accurate simulations of C cycling at locations for which ground-truth data are not available. Thus, this framework enables modellers to simulate current (i.e. last 10 yr) carbon cycling of major agricultural regions. Averaged over the 104 field patches analysed, relative spatial variability for biometry and net ecosystem exchange ranges from ∼7% to ∼18%. The annual sign of net biome productivity is not significantly different from carbon neutrality. (2) Moreover, observing carbon cycling at one single field with its individual sowing pattern is not sufficient to constrain large-scale agroecosystem carbon flux seasonality. Study area average growing season length is 20 days longer than observed at Bondville, primarily because of an earlier estimated start of season. (3) For carbon budgeting, additional information on cropland soil management and belowground carbon cycling has to be considered, as such constraints are not provided by MODIS

The Value of Sentinel-2 Spectral Bands for the Assessment of Winter Wheat Growth and Development

“…This research demonstrates the value of the Sentinel-2 spectral characteristics for retrieving critical variables that can support more sustainable crop management practices. the value of LAI data for updating state variables in process-based agroecosystem models in order to improve estimates of crop yield [3][4][5][6] and land-atmosphere carbon dioxide exchanges [7,8]. On the other hand, chlorophyll is a key driver of plant light absorption and conversion to chemical energy and is, therefore, an indicator of plant health and potential gross primary productivity [9,10].…”

Estimation of Community Land Model parameters for an improved assessment of net carbon fluxes at European sites

“…However, crops are highly diverse in terms of both species grown and management practices applied. Accordingly, previous studies showed that crop parameters are critical to transfer to other sites [ Sus et al, ] or different resolutions [ Iizumi et al, ].…”

Response to Review 2 and proposed changes to manuscript

“…The GPP data does not have an estimate of the uncertainty about that data and so instead we infer the uncertainty associated 10 with that data as σ x,D . In the case of MODIS fAPAR data we explicitly incorporate a measure of variation in the data within the geographical area used to compute the mean fAPAR as well as inferring a parameter representing additional unexplained variation.…”

“…The second data source is flux tower eddy covariance (EC) data which provides high resolution CO 2 fluxes 15 at point locations (Baldocchi and Wilson, 2001). Previously, data assimilation methods have been used for an ecosystem model in croplands with earth observation data (Revill et al, 2013;Sus et al, 2013), but both studies focused on ecosystem carbon fluxes and leaf area index and included no estimates of yield.…”