In developing countries, information on the area and spatial distribution of paddy rice fields is an essential requirement for ensuring food security and facilitating targeted actions of both technical assistance and restoration of degraded production areas. In this study, Sentinel 1 (S1) and Sentinel 2 (S2) imagery was used to map lowland rice crop areas in the Sédhiou region (Senegal) for the 2017, 2018, and 2019 growing seasons using the Random Forest (RF) algorithm. Ground sample datasets were annually collected (416, 455, and 400 samples) for training and testing yearly RF classification. A procedure was preliminarily applied to process S2 scenes and yield a normalized difference vegetation index (NDVI) time series less affected by clouds. A total of 93 predictors were calculated from S2 NDVI time series and S1 vertical transmit–horizontal receive (VH) and vertical transmit–vertical receive (VV) backscatters. Guided regularized random forest (GRRF) was used to deal with the arising multicollinearity and identify the most important predictors. The RF classifier was then applied to the selected predictors. The algorithm predicted the five land cover types present in the test areas, with a maximum accuracy of 87% and kappa coefficient of 0.8 in 2019. The broad land cover maps identified around 12,500 (2017), 13,800 (2018), and 12,800 (2019) ha of lowland rice crops. The study highlighted a partial difficulty of the classifier to distinguish rice from natural herbaceous vegetation (NHV) due to similar temporal patterns and high intra-class variability. Moreover, the results of this investigation indicated that S2-derived predictors provided more valuable information compared to VV and VH backscatter-derived predictors, but a combination of radar and optical imagery always outperformed a classification based on single-sensor inputs. An example is finally provided that illustrates how the maps obtained can be combined with ground observations through a ratio estimator in order to yield a statistically sound prediction of rice area all over the study region.

Grapevine simulation models are mostly used to estimate plant development, growth and yield at plot scale. However, the spatial variability of pedologic and micro-climatic conditions can influence vine growth, leading to a sub-field heterogeneity in plant vigor and final yield that may be better estimated through the assimilation of high spatial resolution data in crop models. In this study, the spatial variability of grapevine intercepted radiation at fruit-set was used as input for a grapevine simulation model to estimate the variability in biomass accumulation and yield in two Tuscan vineyards (Sites A and B). In Site A, the model, forced with intercepted radiation data as derived from the leaf area index (LAI), measured at canopy level in three main vigor areas of the vineyard, provided a satisfactory simulation of the final pruning weight (r² = 0.61; RMSE = 19.86 dry matter g m⁻²). In Site B, Normalized Difference Vegetation Index (NDVI) from Sentinel-2A images was firstly re-scaled to account for canopy fraction cover over the study areas and then used as a proxy for grapevine intercepted radiation for each single pixel. These data were used to drive the grapevine simulation model accounting for spatial variability of plant vigor to reproduce yield variability at pixel scale (r² = 0.47; RMSE = 75.52 dry matter g m⁻²). This study represents the first step towards the realization of a decision tool supporting winegrowers in the selection of the most appropriate agronomic practices for reducing the vine vigor and yield variability at sub-field level.

Conventional meteorological data and remotely sensed Normalized Difference Vegetation Index (NDVI) images can be proficiently combined to predict actual evapotranspiration (ET_A) on different spatial and temporal scales. Up to now, however, the operational application of this approach in heterogeneous Mediterranean regions has found difficulty due to the insufficient spatial resolution of satellite sensors having high acquisition frequency (i.e. 250 m of Terra/Aqua Moderate Resolution Imaging Spectroradiometer, MODIS). The current study investigates the advantages brought for this objective by the recent availability of NDVI data taken from the Sentinel-2 MultiSpectral Instrument (MSI), which has a spatial resolution of 10 m. The investigation has been performed in two Mediterranean areas characterized by different spatio-temporal variability of vegetation cover. The first is a mountain coniferous forest, where such variability is low, while the second is a relatively small (around 10 ha) irrigated tomato field surrounded by other annual crops showing diversified growing cycles. An ET_A estimation method based on NDVI data is applied in the two study areas and its performances are evaluated against ground references obtained through the elaboration of site measurements (i.e. meteorological, soil water content, and crop coefficient observations). The advantage of using MSI over MODIS NDVI images is marginal in the first case study, while is evident in the second. More particularly, such advantage is outstanding when the remote sensing method is applied in an operational mode, i.e. without using the information on the water supplied to the tomato crop by irrigation. This confirms that the utility of higher spatial resolution data is dependent not only on the fragmentation of the observed landscapes but also on the synchronicity of major vegetation growing cycles, which is influenced by both environmental and human-induced factors.

This paper presents the development and testing of different integration procedures aimed at producing Normalized Difference Vegetation Index (NDVI) datasets with high spatial and temporal resolutions. In particular, two methods were considered, which both used higher spatial resolution information to increase the spatial detail of more frequent low resolution NDVI data. The first method was based on the extraction of per‐class NDVI end‐member values from low spatial resolution images (NOAA‐AVHRR), followed by a combination of these values with a reference land cover map. Starting from the same end‐members, the second method introduced further high resolution information by means of regression analyses applied per‐pixel to the multitemporal dataset. The performances of the two methods were evaluated by means of two experiments carried out in a Mediterranean study area (San Rossore, central Italy). First, the images produced were compared on a per‐pixel basis to higher spatial resolution Landsat TM/ETM+ data. Next, the NDVI temporal profiles were transformed into Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) estimates, which were combined with standard meteorological (radiation) data to compute forest Gross Primary Productivity (GPP). The validation of these GPP estimates was performed by comparison with the outputs of a more complex model of bio‐geochemical processes (FOREST‐BGC), already calibrated in the study area. The results of both tests indicated the greater accuracy of the more advanced per‐pixel integration method, which was capable of reproducing NDVI data with high spatial and temporal details.

The estimation of chlorophyll concentration in marine waters is fundamental for a number of scientific and practical purposes. Standard ocean color algorithms applicable to moderate resolution imaging spectroradiometer (MODIS) imagery, such as OC3M and MedOC3, are known to overestimate chlorophyll concentration ([CHL]) in Mediterranean oligotrophic waters. The performances of these algorithms are currently evaluated together with two relatively new algorithms, OC5 and SAM_LT, which make use of more of the spectral information of MODIS data. This evaluation exercise has been carried out using in situ data collected in the North Tyrrhenian and Ligurian Seas during three recent oceanographic campaigns. The four algorithms perform differently in Case 1 and Case 2 waters defined following global and local classification criteria. In particular, the mentioned [CHL] overestimation of OC3M and MedOC3 is not evident for typical Case 1 waters; this overestimation is instead significant in intermediate and Case 2 waters. OC5 and SAM_LT are less sensitive to this problem, and are generally more accurate in Case 2 waters. These results are finally interpreted and discussed in light of a possible operational utilization of the [CHL] estimation methods.