Fractional vegetation cover (FVC) is one of the most critical parameters in monitoring vegetation status. Accurate estimates of FVC are crucial to the use in land surface models. The dimidiate pixel model is the most widely used method for retrieval of FVC. The normalized difference vegetation index (NDVI) of bare soil endmember (NDVI_soil) is usually assumed to be invariant without taking into account the spatial variability of soil backgrounds. Two NDVI_soil determining methods were compared for estimating FVC. The first method used an invariant NDVI_soil for the Northeast China. The second method used the historical minimum NDVI along with information on soil types to estimate NDVI_soil for each soil type. We quantified the influence of variations of NDVI_soil derived from the second method on FVC estimation for each soil type and compared the differences in FVC estimated by these two methods. Analysis shows that the uncertainty in FVC estimation introduced by NDVI_soil variability can exceed 0.1 (root mean square error—RMSE), with the largest errors occurring in vegetation types with low NDVI. NDVI_soil with higher variation causes greater uncertainty on FVC. The difference between the two versions of FVC in Northeast China, is about 0.07 with an RMSE of 0.07. Validation using fine-resolution FVC reference maps shows that the second approach yields better estimates of FVC than using an invariant NDVI_soil value. The accuracy of FVC estimates is improved from 0.1 to 0.07 (RMSE), on average, in the croplands and from 0.04 to 0.03 in the grasslands. Soil backgrounds have impacts not only on NDVI_soil but also on other VI_soil. Further focus will be the selection of optimal vegetation indices and the modeling of the relationships between VI_soil and soil properties for predicting VI_soil.

Quantifying crop residue cover (CRC) on field surfaces is important for monitoring the tillage intensity and promoting sustainable management. Remote-sensing-based techniques have proven practical for determining CRC, however, the methods used are primarily limited to empirical regression based on crop residue indices (CRIs). This study provides a systematic evaluation of empirical regressions and machine learning (ML) algorithms based on their ability to estimate CRC using Sentinel-2 Multispectral Instrument (MSI) data. Unmanned aerial vehicle orthomosaics were used to extracted ground CRC for training Sentinel-2 data-based CRC models. For empirical regression, nine MSI bands, 10 published CRIs, three proposed CRIs, and four mean textural features were evaluated using univariate linear regression. The best performance was obtained by a three-band index calculated using (B2 − B4)/(B2 − B12), with an R²_cv of 0.63 and RMSE_cv of 6.509%, using a 10-fold cross-validation. The methodologies of partial least squares regression (PLSR), artificial neural network (ANN), Gaussian process regression (GPR), support vector regression (SVR), and random forest (RF) were compared with four groups of predictors, including nine MSI bands, 13 CRIs, a combination of MSI bands and mean textural features, and a combination of CRIs and textural features. In general, ML approaches achieved high accuracy. A PLSR model with 13 CRIs and textural features resulted in an accuracy of R²_cv = 0.66 and RMSE_cv = 6.427%. An RF model with predictors of MSI bands and textural features estimated CRC with an R²_cv = 0.61 and RMSE_cv = 6.415%. The estimation was improved by an SVR model with the same input predictors (R²_cv = 0.67, RMSE_cv = 6.343%), followed by a GPR model based on CRIs and textural features. The performance of GPR models was further improved by optimal input variables. A GPR model with six input variables, three MSI bands and three textural features, performed the best, with R²_cv = 0.69 and RMSE_cv = 6.149%. This study provides a reference for estimating CRC from Sentinel-2 imagery using ML approaches. The GPR approach is recommended. A combination of spectral information and textural features leads to an improvement in the retrieval of CRC.

The increasing frequency of fires inhibits the estimation of carbon reserves in boreal forest ecosystems because fires release significant amounts of carbon into the atmosphere through combustion. However, less is known regarding the effects of vegetation succession processes on ecosystem C-flux that follow fires. This paper describes intra- and inter-annual vegetation restoration trajectories via MODIS time-series and Landsat data. The temporal and spatial characteristics of the natural succession were analyzed from 2000 to 2016. Finally, we regressed post-fire MODIS EVI, LST and LSWI values onto GPP and NPP values to identify the main limiting factors during post-fire carbon exchange. The results show immediate variations after the fire event, with EVI and LSWI decreasing by 0.21 and 0.31, respectively, and the LST increasing to 6.89 °C. After this initial variation, subsequent fire-induced variations were significantly smaller; instead, seasonality began governing the change characteristics. The greatest differences in EVI, LST and LSWI were observed in August and September compared to those in other months (0.29, 6.9 and 0.35, respectively), including July, which was the second month after the fire. We estimated the mean EVI recovery periods under different fire intensities (approximately 10, 12 and 16 years): the LST recovery time is one year earlier than that of the EVI. GPP and NPP decreased after the fire by 22–45 g C·m⁻²·month⁻¹ (30–80%) and 0.13–0.35 kg C·m⁻²·year⁻¹ (20–60%), respectively. Excluding the winter period, when no photosynthesis occurred, the correlation between the EVI and GPP was the strongest, and the correlation coefficient varied with the burn intensity. When changes in EVI, LST and LSWI after the fire in the boreal forest were more significant, the severity of the fire determined the magnitude of the changes, and the seasonality aggravated these changes. On the other hand, the seasonality is another important factor that affects vegetation restoration and land-surface energy fluxes in boreal forests. The strong correlations between EVI and GPP/NPP reveal that the C-flux can be simply and directly estimated on a per-pixel basis from EVI data, which can be used to accurately estimate land-surface energy fluxes during vegetation restoration and reduce uncertainties in the estimation of forests’ carbon reserves.

Rapid and accurate estimation of maize biomass is critical for predicting crop productivity. The launched Sentinel-1 (S-1) synthetic aperture radar (SAR) and Sentinel-2 (S-2) missions offer a new opportunity to map biomass. The selection of appropriate response variables is crucial for improving the accuracy of biomass estimation. We developed models from SAR polarization indices, vegetation indices (VIs), and biophysical variables (BPVs) based on gaussian process regression (GPR) and random forest (RF) with feature optimization to retrieve maize biomass in Changchun, Jilin province, Northeastern China. Three new predictors from each type of remote sensing data were proposed based on the correlations to biomass measured in June, July, and August 2018. The results showed that a predictor combined by vertical-horizontal polarization (VV), vertical-horizontal polarization (VH), and the difference of VH and VV (VH-VV) derived from S-1 images of June, July, and August, respectively, with GPR and RF, provided a more accurate estimation of biomass (R² = 0.81–0.83, RMSE = 0.40–0.41 kg/m²) than the models based on single SAR polarization indices or their combinations, or optimized features (R² = 0.04–0.39, RMSE = 0.84–1.08 kg/m²). Among the S-2 VIs, the GPR model using a combination of ratio vegetation index (RVI) of June, normalized different infrared index (NDII) of July, and normalized difference vegetation index (NDVI) of August achieved a result with R² = 0.83 and RMSE = 0.39 kg/m², much better than single VIs or their combination, or optimized features (R² of 0.31–0.77, RMSE of 0.47–0.87 kg/m²). A BPV predictor, combined with leaf chlorophyll content (CAB) in June, canopy water content (CWC) in July, and fractional vegetation cover (FCOVER) in August, with RF, also yielded the highest accuracy (R² = 0.85, RMSE = 0.38 kg/m²) compared to that of single BPVs or their combinations, or optimized subset. Overall, the three combined predictors were found to be significant contributors to improving the estimation accuracy of biomass with GPR and RF methods. This study clearly sheds new insights on the application of S-1 and S-2 data on maize biomass modeling.

Fractional vegetation cover (FVC) is an important variable for describing the quality and changes of vegetation in terrestrial ecosystems. The simplest and most widely used model for the estimation of FVC is the dimidiate pixel model. The normalized difference vegetation index (NDVI) is commonly used as a vegetation index (VI) in this model. A range of VIs is possible alternative to the use of NDVI in the dimidiate pixel model. In this article, six VI-based dimidiate pixel models were compared using in situ measurements and canopy reflectances simulated by the PROSAIL model over nine different soil backgrounds. A comparison with in situ measurements showed that the Gutman–Ignatov method overestimated FVC, with a mean root mean square error (RMSE) of 0.14. The mean RMSE had an intermediate value of 0.08 in the Carlson–Ripley method and was further reduced to 0.05 in the method proposed by Baret et al. The use of both modified soil-adjusted vegetation index (MSAVI) and a mixture of NDVI and the ratio vegetation index (RVI) to replace NDVI in the Gutman–Ignatov model reduced the RMSE to 0.06. The mean RMSE in the difference vegetation index (DVI)-based model was 0.08. The simulated results indicated that soil backgrounds have significant effects on these VI-based models. The sensitivity of the first three models and the NDVI plus RVI-based model to soil backgrounds decreased with an increase in soil reflectance. In contrast, the DVI-based model is sensitive to soil backgrounds with high reflectances. MSAVI, which is less sensitive to soil backgrounds, represents a feasible alternative to the use of NDVI in the Gutman–Ignatov model.