Machine learning estimates of eddy covariance carbon flux in a scrub in the Mexican highland

Abstract

Abstract. Arid and semiarid ecosystems contain relatively high species diversity and are subject to intense use, in particular extensive cattle grazing, which has favored the expansion and encroachment of perennial thorny shrubs into the grasslands, thus decreasing the value of the rangeland. However, these environments have been shown to positively impact global carbon dynamics. Machine learning and remote sensing have enhanced our knowledge about carbon dynamics, but they need to be further developed and adapted to particular analysis. We measured the net ecosystem exchange (NEE) of C with the eddy covariance (EC) method and estimated gross primary production (GPP) in a thorny scrub at Bernal in Mexico. We tested the agreement between EC estimates and remotely sensed GPP estimates from the Moderate Resolution Imaging Spectroradiometer (MODIS), and also with two alternative modeling methods: ordinary-least-squares (OLS) regression and ensembles of machine learning algorithms (EMLs). The variables used as predictors were MODIS spectral bands, vegetation indices and products, and gridded environmental variables. The Bernal site was a carbon sink even though it was overgrazed, the average NEE during 15 months of 2017 and 2018 was −0.78 gCm-2d-1, and the flux was negative or neutral during the measured months. The probability of agreement (θs) represented the agreement between observed and estimated values of GPP across the range of measurement. According to the mean value of θs, agreement was higher for the EML (0.6) followed by OLS (0.5) and then MODIS (0.24). This graphic metric was more informative than r2 (0.98, 0.67, 0.58, respectively) to evaluate the model performance. This was particularly true for MODIS because the maximum θs of 4.3 was for measurements of 0.8 gCm-2d-1 and then decreased steadily below 1 θs for measurements above 6.5 gCm-2d-1 for this scrub vegetation. In the case of EML and OLS, the θs was stable across the range of measurement. We used an EML for the Ameriflux site US-SRM, which is similar in vegetation and climate, to predict GPP at Bernal, but θs was low (0.16), indicating the local specificity of this model. Although cacti were an important component of the vegetation, the nighttime flux was characterized by positive NEE, suggesting that the photosynthetic dark-cycle flux of cacti was lower than ecosystem respiration. The discrepancy between MODIS and EC GPP estimates stresses the need to understand the limitations of both methods.

Monitoring Energy Balance, Turbulent Flux Partitioning, Evapotranspiration and Biophysical Parameters of Nopalea cochenillifera (Cactaceae) in the Brazilian Semi-Arid Environment

“…It is possible that the proximity of turbulent fluxes in the morning (before 10:00) during some wet and dry seasons may have been due to the increased atmospheric demand and soil moisture that favour evapotranspiration. CAM plants have the ability to cool the soil overnight [43], and although their metabolism is generally nocturnal, we believe that this thermal reduction lasted until the early hours of the morning [27]. In this case, evaporation from the soil may have resulted in greater change in latent heat flux.…”

“…To quantify the land-atmosphere fluxes, and understand this partition, the surface energy balance (SEB) predicts variations in turbulent fluxes and evapotranspiration (ET) from interaction of the soil-vegetation-atmosphere system [23][24][25]. In addition to being considered a fairly robust and valid method under semi-arid conditions, it can be applied to different vegetated surfaces (e.g., areas of forest and agricultural crops) and areas with small footprints (i.e., different fetch-to-height ratios) [26][27][28]. Among the methods used to determine turbulent flux, the Bowen ratio indirectly quantifies the latent heat flux (LE) using net radiation (R n ), soil heat flux (G) and the air temperature and humidity gradients [29,30].…”

Evaluation of Selected Sentinel‐2 Remotely Sensed Vegetation Indices and MODIS GPP in Representing Productivity in Semi‐Arid South African Ecosystems

“…This is consistent with other sites where GPP MODIS generally underestimated GPP EC by about 34% across 15 Fluxnet sites (L. Wang et al., 2017), with strong underestimation also observed in semi‐arid ecosystems in the Sahel (Tagesson et al., 2017) where about 67% variability of GPP was explained by GPP MODIS . About 60% of GPP EC variability was explained by GPP MODIS at scrub site in the Mexican highland (Guevara‐Escobar et al., 2021). These are within range with Ben_Sav and Ben_Kar (68%–70%), while at Mid_Kar (83%), GPP MODIS performed considerably better.…”

Challenges in studying water fluxes within the soil-plant-atmosphere continuum: A tracer-based perspective on pathways to progress

“…Recently, machine learning and artificial intelligence methods have been utilized in the eddy-covariance and, to a lesser degree, gas-exchange chamber flux community. Starting with large and high resolution eddy co-variance data (Guevara-Escobar et al, 2021;Zhu et al, 2022), there have been recent…”