**4. Discussion**

In this paper, we presented the correlation between the ECI (based on emissivity data from infrared spectral channels of the CAMEL dataset) and the WDI (based on the data from ECMWF analyses) against ET measured in four EC sites in Italy (Renon, Monte Bondone, Lavarone and Bosco della Fontana). Furthermore, the behavior of the above mentioned correlations in relation to the vegetation cover was discussed. Additionally, the environmental heterogeneity was assessed (using NDVI MODIS images and analyzed through the SH Rao's Q index) in order to evaluate the performance, particularly of the ECI in relation to the habitat fragmentation over the four study areas.

The results show that the correlation between ECI and ET is statistically significant for the four study areas. In the areas of Renon, Lavarone and Monte Bondone, the relationships are positive and they are influenced by the snow cover (estimated with the "snow fraction" layer of the CAMEL dataset), which interferes with the emissivity values. In the winter months, when the snow covers the earth surface, the ECI yields a second peak that does not follow the ET trend. As described by Masiello et al. [29,35], the ECI is highly influenced by the dryness and by the land cover. As an example, the ECI has a value of around 0.98 for water-covered surfaces (e.g., snow), whereas, for dry and senescent vegetation, the index reaches lower values. The ECI was initially developed to build synergy with the NDVI to overcome the drawback of the vegetation index that, in certain cases (as in the case of Masiello et al. [35] over the Congo basin area for a temporal range of seventeen years), is not able to discriminate senescent vegetation and bare soil; in particular, when the vegetation regeneration start after a deforestation or degradation event [35]. On the other hand, in the area of Bosco della Fonatana, the correlation ECI–ET was negative: in the considered year, the curves indeed had an opposite trend. Several reasons can explain this different trend: since the ECI is influenced by the dryness and by the land cover, its values could be distorted by the humidity of the soil and by the presence of surface water, which is very high in this area, located near the aquatic ecosystems of the Mincio river and of the Mantova's lakes. The other reason is related to the high habitat fragmentation within the CAMEL pixel that alters the ECI value (as shown in in Figure 7) and summarized in Table 1. The results of the table indeed show that the Rao's Q index, used to estimate the environmental heterogeneity (by the assessment of the SH), has the highest value in the area of Bosco della Fontana.

The concept of the SVH and the related environmental heterogeneity is therefore crucial not only in the estimation of biodiversity and the assessment of species diversity [41,61,62], but also in all studies where the ground information (e.g., energy fluxes) is correlated with the remote sensing data. This aspect is particularly common when using remote sensing information. An image, through the diversity of the various pixels, provides knowledge about the remote sensing response caused by the physical interaction of the measured information (emissivity in our case) with the Earth's surface. Thus, the raster grid of pixels that build up an image, represents just an average response of a real information. The question regarding the size of the pixels, in order to describe a certain area or to characterize a real situation or an event is, as previously stated, still debated in the research community [63,64]. Like in our case, images with a coarse spatial resolution tend to integrate the information of various subjects (e.g., vegetation, human artefacts, rivers.), homogenizing the signal and causing difficulties toward clearly identifying boundaries between spatial entities (individuals, vegetation types, ecosystem types) [65]. On the other hand, a fine spatial resolution may lead to a level of details within spatial entities that may cause a strong heterogeneity, leading to strong noises and uncertainties [66].

As far as the relationship of ET vs. WDI is concerned (Figures 5 and 6), we have found that the correlation is less sensitive to the vegetation changes and environmental heterogeneity. The slope of the linear regression remains positive, although we have explored sites with a very large heterogeneity. The difference in temperature (Ts*ECMWF* — Td*ECMWF*) is correlated with the ET because the surface temperature is strongly dependent on the impinging solar radiation, whereas the dew point temperature is dependent on both the air temperature and the humidity field. The partial mismatch between the two variables might be due to the lack of an additional meteorological parameter, such as the wind. The detail of the regression still seems to be site-dependent; however, in this case, there is no ambiguity that ET and the WDI are positive correlated. Since the WDI is based on temperatures, it is less influenced by the habitat fragmentation and by the environmental heterogeneity compared to the ECI, which is based on the emissivity of the surface. For this reason, the correlation holds true also in the area of Bosco della Fontana, which, as previously stated, showed an opposite result in the correlation ET-ECI.

Finally, we believe that the synergistic use of the ECI and WDI might increase the accuracy of ET estimation because of their different sensitivities to different aspects of the vegetation. The ECI is better suited to detecting changes in the vegetation state, green to senescent or transition to bare soil. These states can influence ET, but in a way that can be highly nonlinear. Conversely, the WDI is more linearly related to ET. The synergistic use of the two could be, e.g., of some interest during an intense heatwave, which has become common in temperate regions because of climate change (e.g., see https://climate.copernicus.eu/esotc/2021/globe-in-2021 (accessed on 8 May 2012)). In the event of heatwaves, we expect the WDI > 0 (e.g., see [32]), and a decreasing value of the ECI could show an early decay of the foliage to the senescent state, and hence vegetation stress.
