Mapping Apennines river paths along different hydrological conditions from satellite images: A description of the method and potential applications

Climate change is producing large impacts on rivers, amplifying hydrological extremes. Prolonged drought periods result in dramatic stress for river biota and associated processes due to low discharge, reducing the interactions between rivers and their lateral environments or leading to hydrological intermittency. New quantitative methods are needed, to correlate discharge with the available riverine habitats. In this work we have mapped the wet surface and paths of two stretches of the Taro and Trebbia Apennine rivers, analyzing satellite images from periods with contrasting discharge. The considered stretches are critical due to different human pressures (large water withdrawals for agriculture and industrial use) and are particularly vulnerable to further, climate‐driven discharge reductions. The produced images offer multiple possibilities to extract qualitative and quantitative information at the whole stretch scales, including habitat reduction along with decreasing discharge, threshold discharge limiting lateral interactions, or the evaluation of longitudinal river continuity. We discuss the limitation and the potentialities of the method and the maps produced in terms of possible application in the field of river geomorphology, ecology, the definition of ecological river flow, risk assessment, and river management.


| INTRODUCTION
Unusually high temperatures, water scarcity, and prolonged drought events, even during winter months, are among the most pressing environmental challenges of the 21st century (Tramblay et al., 2020;Vorosmarty et al., 2000).Such threats may lead to pronounced changes in river discharge and in the timing of flood and lean events, sometimes resulting in hydrological intermittency (Gampe et al., 2016;Sarremejane et al., 2022).These projections affect not only Mediterranean and arid regions but also streams and rivers in temperate areas, where extended periods of desiccation have been observed (Sutherland et al., 2008;Wilby et al., 2006).In recent decades, the intensity and frequency of droughts have increased dramatically across Europe, due to global climate change and enhanced anthropic pressures causing the shift of perennial streams into systems with very low to intermittent flow regimes, with effects on many organisms, and on abiotic and biotic processes within lotic ecosystems (Doretto et al., 2018;Lange & Haensler, 2012;Piano et al., 2019).
Flow is a key driver of the structure and function of aquatic ecosystems (e.g., Poff & Zimmerman, 2010;Vannote et al., 1980) and its marked reduction has pervasive ecological effects on freshwater environments (Larned et al., 2010).As river biological diversity and functioning depend on longitudinal water transport, the reduction of water discharge and velocity alters a wide range of physical parameters, including the rates of gas exchange, the river-groundwater mixing ratios, and the surface and heterogeneity of available aquatic habitats, producing a cascade of consequences on the biota.During droughts, decreased water flow and dilution capacity increase nutrient availability, favor the growth of primary producers, and reduce the export of fine particulate organic matter downstream, resulting in the deposition on the riverbed.Indeed, extensive macroalgal growth often occurs in rivers and streams undergoing pronounced flow reduction (Mosley, 2015).If during the early growth phase of macroalgae large oxygen production might be beneficial for aquatic organisms, the subsequent biomass accumulation results in increased respiration of the benthic system, and clogging phenomena impede the water exchange with the riverbed.
The hyporheic zone is a principal component of stream ecosystems (Grimm & Fisher, 1984;Jones & Holmes, 1996;Krause et al., 2011) and because of the high exchange of water and the high activity of microbes, it plays a key role as a reactive zone, promoting high turnover rates of organic matter and the exchange of oxygen and nutrients, as well as providing a habitat for benthic communities and larval stages of aquatic vertebrates (Boulton, 1999;McClain et al., 2003).Flow reduction and fine sediment accumulation have been shown to lead to changes in microbial biomass, community composition, biogeochemical processes, and water chemistry of the hyporheic zone (Amalfitano et al., 2008;Doretto et al., 2016;Fierer et al., 2007;Rees et al., 2006;Zoppini et al., 2010).Oxygen shortage, minor dilution or accumulation of metabolic end-products, and an increasing proportion of anaerobic microbial processes may result in consequences.
As compared to limnology, river ecology is a young discipline that has grown in recent years and is accumulating knowledge on the factors regulating the functioning of river ecosystems.Indeed, there is an urgent need to understand the effects that climate change, reduced water availability, and hydrological intermittency on river ecology.
Studies focus on the dynamics of macrofauna and on the implications for organic matter processing and downstream nutrient transport (Corti et al., 2011;Palmia et al., 2019).They also analyze the possible implications for river groundwater interactions, hatch success of crustaceans, fish, and amphibians using riverbed for eggs deposition and sediment transport (Uys & O'keeffe, 1997;Vander Vorste et al., 2020;Whiles & Goldowitz, 2001).Many of these studies are local or localized in small portions of river courses and do not take into account wider geographical scales (the river as a whole).
Small river ecosystems tend to be particularly variable over short spatial and temporal scales, which makes it very difficult to have continuous information through traditional surveys (Allen et al., 2018;Lu et al., 2020).For this reason, satellite and/or airborne images are increasingly used for mapping rivers (e.g., McCabe et al., 2017;Piégay et al., 2019;Yamazaki et al., 2015) up to producing global data sets at 30 meters scale with Landsat satellite data (e.g., Allen & Pavelsky, 2015;Gleason & Smith, 2014).
In addition to the Landsat archive, the images of MSI sensors onboard Sentinel-2 A/B satellites have been available since 2015.
In this paper, we present an approach based on a large spatial and temporal scale analysis of two Apennine rivers (Taro and Trebbia, northern Italy) that display large discharge variations due to combined human water use (irrigation) and climate change.A large number of satellite images (21) were collected in river stretches within Nature 2000 sites (Directive 92/43/CEE) and processed in order to produce maps of the watercourses under various hydrological conditions, including hydrological extremes.We discuss multiple possible uses of the information gathered from these maps, that are of potential interest to different stakeholders, including river ecologists, conservation biologists, Natural Parks, irrigation consortia, or companies responsible for sewage treatment plants.Such maps can also support the experimental definition of ecological flows in the implementation of the Water Framework Directive (European Commission, Guidance No. 31, 2015).

| STUDY AREA
The Taro and Trebbia rivers are located on the north-eastern flank of the Northern Apennines (Northern Italy) and their catchments cover an area of about 2026 and 1070 km 2 , respectively (Figure 1).The physiographic features of both catchments consist of mountainous and hilly areas for the Taro (77% of the total) and for the Trebbia (85% of the total) catchments, with a maximum relief of about 1735 and 1406 m a.s.l., respectively.Both rivers have their terminal reaches in the Po plain, between the Apennines and the confluence with the Po River.The geological features of both catchments are similar.
The mountain area of both river courses is characterized by sedimentary rocks, mainly limestones, marls, sandstones, claystones, and the outcroppings of ophiolitic rocks.The outcropping rocks are in general highly tectonized as a result of the long and complex tectonic history of the northern Apennines (Conti et al., 2020).The catchment portions in the Po plain, are formed by successions of marine and continental deposits, Plio-Quaternary in age (Ori, 1993).
The climate is characterized by a cold winter and a dry summer season; the mean annual rainfall for the mountain sectors is 1260 mm yr À1 for the River Taro and 1440 mm yr À1 for the Trebbia River (ARPAE, 2020a,b), with most of the precipitation occurring during autumn and spring, with October and April being, in general, the wettest months.
Discharge data for the two rivers are limited to a few stations.
For the Trebbia River, discharge data are available at Rivergaro gauge station (148 m a.s.l.; 1 in Figure 1); here, the mean annual discharge for 2020 and for the period 2003-2019 was 20 and 23.6 m 3 s À1 , respectively.For the Taro River, discharge data are available at the gauge station of San Secondo Parmense (42 m a.s.l.; 2 in Figure 1); here, the mean annual discharge for 2020 and of the period 2006-2019 was 30.5 m 3 s À1 and 40.6 m 3 s À1 , respectively (Arpa Emilia-Romagna, 2020; Annali Idrologici Arpa Emilia Romagna, 2020).
The spatial pattern of the Trebbia River channel morphology is strongly controlled by the physiographic conditions of the valley, with frequently confined meanders in the upper reach, followed by confined reaches crossing the hilly areas, and then unconfined reaches with a tendency toward braiding along a wide alluvial fan included in the Po River plain (Bollati et al., 2014).
The characteristics and evolution of the Taro River appear strongly conditioned by the structural setting of the area.In fact, the course of the stream coincides with a great fault, whose activity caused the uplifting of the western side by 10-30 meters on average during the Plio-Quaternary (Bernini & Papani, 1987).During this time, the river channel progressively shifted eastward, as testified by the presence (in the middle and low plains) of paleochannel tracks almost exclusively on the western side of the present river course.
Regarding the channel pattern, the unconfined reach is typically braided in the upper part for about 19 km, while it is single-thread mainly meandering in the lower part for about 29 km.The middle reach of about 6 km is transitional, with wandering a character (Clerici et al., 2015).
From an ecological point of view, both rivers belong to the socalled "Continental bioregion" delimited by the southern alpine border, the eastern Apennines, and the Adriatic Sea (https://ec.europa.
eu/environment/nature/natura2000/biogeog_regions/; accessed on May 10th, 2022).In this work, we focused on two specific segments of the Taro and Trebbia rivers (A and B, respectively in Figure 1

| OPTICAL SATELLITE IMAGES ANALYSIS FOR THE RIVERS' EVOLUTION MONITORING
To study the morphodynamical evolution of the Taro and Trebbia rivers and any changes occurring in their surrounding areas, a multitemporal approach was carried out in the two areas of interest.The pixels in the scene, and to distinguish vegetation from bare soil, which are pixels with different colors in the same scene (blue and yellow pixels, respectively, in the FCC carried out in this work).From 2016 to 2020, "synoptical" tables are containing the IR band 8 and FCC images covering the spring-autumn interval.Every table (Figure 4) is a helpful tool to highlight differences that might occur within the considered group of scenes (e.g., changes in pixel color distribution and frequency) and that should correspond to changes that have occurred in the landscape (e.g., seasonal river flow or land use) over a year.In addition, the simultaneous comparison of synoptical tables, each corresponding to a different year, provides the opportunity to highlight long-term changes, that is, over many years.Furthermore, focusing on the river dynamic analysis, to enable the multitemporal overlay of the river paths for different time intervals, a shapefile was extracted within the GIS environment for each satellite image representative of the "river's wet path" (Figure 5).The "wet path" corresponds to those dark-to-black pixels in the FCC images that belong specifically to the floodplain area and correspond to the actual path of an active river at a certain time.The extracted shapefiles, in addition to allowing multiple overlaps of rivers path only, also allow a quantitative value to be assigned to each pixel of the path, related to different river characteristics such as chemical composition or bathymetry.In this regard, bathymetric values were extracted from the same Sentinel 2 MSI dataset for the two extremes of the spring-autumn time span (i.e., March/April and September/October) as described in the following section and then associated for each year, from 2016 to 2020, to the "river's wet path" shapefile.According to a multitemporal approach, to quantify the possible changes occurring in the landscape highlighted by the FCC images comparison over the considered time interval, a supervised classification was carried out on each image using the ENVI (Environment for Visualizing Image) software.The image classification was based on the "Maximum Likelihood" algorithm that considers four regions of interest (ROI) identified in each satellite image and corresponding to water, bare soil, vegetation, and soil-and-vegetation.To evaluate changes that could have affected the landscape, the percentage distribution of the four classes (i.e., water, bare soil, vegetation, and soil-and-vegetation) was calculated for each image and the values were compared over the time interval considered.

| RETRIEVED RIVER BATHYMETRY FROM SATELLITE IMAGE ANALYSIS
A variety of algorithms have been developed for remote sensing of river bathymetry from passive optical image data (Legleiter & Harrison, 2019).This includes depth retrieval via simple linear regression (e.g., Winterbottom & Gilvear, 1997) techniques based on hydraulic principles (e.g., Legleiter, 2015), probability concepts for the Taro River consisting of Infrared band 8 (upper) and false colors compositions (FCC) of Infrared bands 8a, 11, 12 from Sentinel 2 MSI imagery, covering the 2016 spring-autumn interval.The "synoptical table" highlights river changes along with variations in the adjacent floodplain areas (i.e., yellow and blue polygons) over the considered time interval.[Color figure can be viewed at wileyonlinelibrary.com] (Legleiter, 2016), or radiative transfer modeling (e.g., Kerr & Purkis, 2018).To retrieve the bottom depth of the river, it is necessary to establish a relationship between reflectance and water depth considering that the signal emerging from the water depends not only on depth but also on the reflectance of the streambed, the optical properties of the water column, and any light reflected from the water surface (Legleiter et al., 2004).For this reason, all the images used in this research to retrieve river bathymetry were previously atmospherically corrected to remove all the noise and disturbance due to the atmosphere.Atmospheric correction was carried out using the 6SV code (Second Simulation of the Satellite Signal in the Solar Spectrum -Vector, Vermote et al., 1997).The code was parametrized with the Continental aerosol model, the AOT at 550 nm and water vapor values were retrieved from daily MODIS products and Ozone concentration from OMI-Aura (Ozone Monitoring Instrument), via NASA Giovanni interface (Acker & Leptoukh, 2007).A bio-optical model was thus applied to atmospherically corrected images to obtain bathymetry maps through the BOMBER (Bio-Optical Model-Based tool for Estimating water quality and bottom properties from Remote sensing images, Giardino et al., 2012) tool.Bottom depth was achieved with the bio-optical model implemented in BOMBER.The model was parameterized using specific inherent optical properties of clear waters with starting values typically of our case study; in particular, chlorophyll was set at 1 mg m À3 , total suspended matter at 0.8 mg m À3 , and colored dissolved organic matter at 0.05 m À1 .The bottom albedo value, at a reference wavelength, was expressed as a linear combination of three substrates having different albedo (sand, rocks, and macrophytes vegetation).

| Rivers evolution from multitemporal analysis of Sentinel 2 imagery
The infrared FCCs images derived from the Sentinel 2 MSI selected data allowed clear distinction among water, vegetation, and bare soil in the study areas, as they correspond to dark-to-black, blue, and yellow pixels in the scenes, respectively.Consequently, the multitemporal image analyses carried out for the spring-autumn interval of each year, from 2016 to 2020, were able to highlight how the landscape has undergone significant changes over time.These changes were essentially related to the variations of the Taro and Trebbia wandering channels' shape, size, and number and to the land-use F I G U R E 5 Different "wet paths" (in yellow, orange, light blue, and purple) for the Taro (upper) and Trebbia (lower) rivers extracted as shapefile (multi-points) within the GIS environment from Infrared bands of Sentinel 2 MSI imagery, for the spring-autumn time interval.The "wet path" corresponds to the dark-to-black pixels in the Infrared bands that draw the actual path of the active channels.The "wet paths" allow quantitative data on chemical composition or bathymetry to be assigned to each point, enabling easy and interactive comparison to identify changes, qualitatively and quantitatively.[Color figure can be viewed at wileyonlinelibrary.com] modifications occurring in the adjacent alluvial areas.Regarding changes that directly affect rivers, it was observed that generally, channels tend to thin out and sometimes disappear from springtime to the beginning of the autumn, with a consequent reduction in the number of active channels crossing the floodplain.Although this can be considered a general and expected trend following the timing and intensity of precipitation, the image analyses reveal that some reaches of the Taro and Trebbia rivers seem to have been more affected by these changes than others.In the middle portion of the Taro River, which is part of a natural park and was characterized by a wide river floodplain area usually hosting many active channels, the variations appear most evident (Figure 6,upper).This is probably due to water withdrawals for irrigation that further amplify water scarcity during summer, and whose effects can be quantified in terms of habitat reduction.Similarly, in the northern tract of the Trebbia River, which is characterized by a larger river valley crossed by wandering channels differently from the southern portion that is more a meandering channel, changes are most noticeable (Figure 6, lower).Comparing changes affecting both rivers in the spring-autumn time interval over 5 years (i.e., 2016-2020), it appears that the trend just described generally characterizes all the time span considered.However, within the 5-year observation interval, it is possible to recognize some peculiarities.The general tendency of wandering channels to thin and sometimes disappear over the spring-autumn time span appears especially prominent in 2017 while it is less clearly identifiable in 2020, both for the Taro and Trebbia rivers (Figure 7a,b).This is likely due to the large interannual heterogeneity of summer precipitation events, that were very close to historical minima and maxima during 2017 and 2020, respectively.The availability of high-quality satellite images, such as commercial images, would allow rapid and accurate evaluations, carried out at the mesoscale, of the relationships between discharge and the availability of habitats.Such evaluations would support the definition of ecological flow in different river stretches and would allow the identification of critical areas where the loss of habitat for fish and macrofauna is particularly severe during summer.To complement the results derived from the image analysis on channels' morphologic evolution, an assessment of possible variations in channels' depth was carried out.The "bathymetric" values, intended as water column depth (in case of the presence of macrophytes the values are the canopy and not the bottom), extracted from satellite images of each year, were generally less than 1 m along the river path with the exception of some short reaches where values are around 5 m in the case of the Taro River, while it is generally greater than 1 m in the case of the Trebbia River.However, the depth of the water column, both for the Taro and Trebbia rivers, tends to increase along the whole river course during the springtime reaching mean values around 1.5 m and more than 2 m in the Taro and Trebbia rivers, respectively, otherwise reaching the minimum, around 1 m or less, in both cases during the summer (Figure 8).River bathymetry revealed with satellite remote sensing can be affected by different sources of errors such as georeferencing errors (Dietrich, 2017), atmospheric correction (Goodman et al., 2008), parametrization of the algorithms (Casal et al., 2020) and mixed pixel size of satellite data (Hedley et al., 2018).However, as interpretation errors are the same in different periods of the year, the relative difference between spring and summer suggests a generalized major reduction of wet river reaches in the two systems.Elsewhere,

| Satellite images analysis: Potentials and limitations
The current research highlights that Sentinel MSI images represent a precious archive of free-of-charge surface data with a high-frequency revisiting time, which is particularly useful to investigate rivers evolution.Satellite data represents the possibility to study rivers and the surrounding areas at different temporal (i.e., days and/or years) and spatial scales (i.e., regional and/or local scales) and collect with a single acquisition information on different surface features both from a qualitative and quantitative point of view.In our case, in fact, it was possible to clearly identify rivers and detect changes affecting their shape and size over the selected time interval, as well as to highlight variations in the land use occurring in the surrounding areas, providing the chance to study the possible relationship between landscape variations and human activity.However, a limitation to be considered in using Sentinel MSI data in such an investigation is the spatial resolution of the satellite imagery (i.e., pixel size) with respect to the dimensions of the studied feature.The pixel size represents a detection limit and, as in our case, it can be often nearly of the same order, or slightly below, of the investigated features thus preventing precise quantitative estimations.A low-medium spatial resolution, as in the case of Sentinel 2 MSI imagery, can be locally responsible for ambiguity during the attribution process (i.e., image classification) of those pixels containing at the same time two different features (i.e., bar soil and vegetation) to a specific feature class.These limitations, however, can be overcome using images from commercial satellite archives, that offer data with spatial resolution around a meter (e.g., WorldView-3), allowing more accurate results to be obtained in terms of spatial resolution and, hence, possible quantitative estimations of surface elements (e.g., length and area) even though they might have higher uncertainty in terms of radiometric resolution.Imagery acquired by drones can also represents a potential with their excellent geometric resolution (even less than 5 cm) for studying rivers evolutions (Ansari et al., 2021;Gracchi et al., 2021), however, the extent of the area investigated and the revisit time for multitemporal analysis could be limitations for their use.Covering a wide area such as the case study, in fact, may be time-consuming and, at times, may be difficult due to the need for drone flight permits.Furthermore, the multi-temporal analysis over several years would hardly be possible with drone images or not feasible if the analysis involves past years, as is the case in the present research.

| Ecological implications and applications
The outcomes of the present research are multiple and can support different multidisciplinary investigations including studies on the combined effects of anthropogenic water use and climate change, the suitability of different river reaches for fish reintroduction and restocking, the definition of minimum or optimal ecological flows, the planning of water resource management and exploitation in the context of future water availability, the understanding of river functioning with respect to variable discharge, the assessment of riparian vegetation and habitat modifications, and the analysis of flood or drought risk scenarios.

| Water chemistry
Different sampling activities carried out under different hydrological conditions reveal that water chemistry does not reflect the ecological status of the river under low or very low discharge.Chemical parameters such as dissolved oxygen or nutrient concentrations did not display large variations, proportional to the variations of discharge.
Dissolved reactive phosphorus, ammonium, and seasonal chlorophyll concentrations for example were always below 0.5 μM, 2 μM and 5 μg L À1 in the two rivers and did not show significant temporal trends (Bartoli, unpublished data).These results either suggest the absence of significant diffuse or point pollution sources affecting water quality during low discharge periods or the occurrence of internal mechanisms (e.g., higher growth of benthic algae during low discharge and higher uptake of nutrients, contrasting point or diffuse inputs, and minor dilution).Elsewhere, a dramatic reduction of river discharge in reaches crossing areas of river-groundwater interactions results in a marked increase of nitrate, due to the replacement of river water with nitrate-polluted groundwater (Delconte et al., 2014;Racchetti et al., 2019).In the Taro and Trebbia systems the multiple anthropogenic pressures (e.g., agriculture, land use, wastewater treatment plants) seem, therefore, to produce limited effects on water chemistry and outputs from chemical analyses alone might lead to an overestimation of the present ecological status of these rivers.Moreover, chemical analyses are generally carried out over a limited number of stations and provide very limited spatial information.On the contrary, morphometric parameters, the diversity, and the number of habitats available to macroorganisms and microorganisms are dramatically affected by the pronounced discharge variation and need to be included in ecological evaluations of these rivers.A tool such as satellite images that allows quick and frequent calculation of wet river areas, isolated reaches or pools, or a list of habitats for different fish communities would appear to be very useful.

| Fish management
The products that can be obtained with the methodologic approach outlined in the present include the inventory of reaches or spots where water pools remain even under the lowest discharge, following prolonged drought periods.These spots can be quantified in terms of location, total surface, and depth.The identification of such spots can be useful to select sampling areas for various purposes including the demographic assessment of fish populations which is fundamental to calculating the fish index requested by the Water Framework Directive (Pagani et al., 2021) or to define various Fish Suitability Indexes.
Similarly, the identification of hydrometric variability based on seasonal differences may direct restocking programs or better define management plans for the conservation of threatened species or

| Soil system budgets, crops, and riparian buffers
Together with the measurement of the wet riverine surfaces where biogeochemical transformations can occur, the seasonal analysis of the land cover or of the distribution and activity of terrestrial primary producers, including cultivated plants, may help our understanding of the seasonal nutrient budget at the whole watershed level.Nutrient budgets are generally performed over an annual period and there is large uncertainty in the mechanisms linking seasonal nutrient genesis, transport, and transformations from the land to the water (Pinardi et al., 2022).As in modern agriculture, the vegetative period of crops is shorter and shorter, maps reporting the seasonal distribution of active vegetation and plowed soil can help understand the mechanisms leading to soil erosion and sediment transport to rivers, identifying therefore vulnerable areas to both diffuse dissolved and diffuse particulate river pollution.

| CONCLUSIONS
This study reports examples of outcomes of satellite image analysis of two river stretches undergoing pronounced seasonality in terms of discharge, wet surface, water depth but also land use in adjacent terrestrial areas.It was possible to observe that in general, from springtime to early autumn, the multiple channels of Taro and Trebbia rivers tend to reduce in size and depth and, hence, sometimes to disappear, with a consequent reduction in the number of active channels.
In addition, image analyses revealed that some portions of the Taro and Trebbia investigated seem to have been affected by these changes more than others, i.e., the wandering channels portions where withdrawal works were located.The areas adjacent to the rivers, which are used for agriculture, also seem to change from spring to autumn, with large variations in the ratio of vegetated to bare soils.It is therefore evident how maps produced from the analysis of satellite imagery can add important and holistic information to scarce, spatially, and temporally limited and fragmented river datasets.We discuss the multiple potential uses of these products in river ecology, river restoration, river and fish management, and biogeochemistry.They provide an alternative approach to the traditional analysis of riverine systems and allow the effects of multiple water use, climate, and land use changes on river diversity and functioning to address.Satellite-derived products are expected to increase in terms of quality and availability, and modern river ecology will surely benefit from such information.
More calibration work is needed to link river discharge, depth, dissolved and particulate nutrient loads, biological diversity, and functioning with satellite-derived products before such products become predictive tools that can be fully integrated into river research.The increasing occurrence of hydrological extremes and the effects they produce on rivers should stimulate further research in this direction.
), corresponding to their terminal intra-valley portion up to the outlet onto the Po Plain, which comprises 11 sites for the ecological monitoring, 6 for the Taro River and 5 for the Trebbia River respectively (Figure 2, upper).In detail, the investigated reach of the Taro River is around 10 km long and extends from Rubbiano (SW) to Medesano (NE) villages, while in the case of the Trebbia River, the reach is around 9 km long and extends from south of Rivergaro (S) to Canneto Sotto (N) villages.(Figure 2, lower).
spring-autumn interval was considered as the observation period for a time span of 5 years, from 2016 to 2020.For this time interval, 21 multispectral optical satellite images (Figure3) from Sentinel 2 MSI (MultiSpectral Instruments) covering the Taro and Trebbia rivers study areas were acquired from the ESA-Copernicus website (https:// scihub.copernicus.eu/dhus/#/home).Data from Sentinel 2 MSI, both radiance and reflectance imagery, were selected considering only the scenes with the minimum percent of cloud cover since the cloud cover represents a disturbance element for image analysis.Infrared (IR) spectral bands from the imagery, which are least affected by atmospheric effects, were combined to obtain false color composite images (FCC) useful for highlighting and discriminating different objects of various natures characterizing the scenes.In particular, the use of infrared bands for the FCC enables the detection, at the same time, of water occurrence on the ground surface, that is dark-to-black F I G U R E 1 Study area location (red box).The investigated reaches of the Taro and Trebbia rivers are highlighted with the yellow rectangles A and B, respectively.(1) Rivergaro gauge station in the Trebbia River; (2) San Secondo Parmense gauge station in the Taro River.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 2 (Upper) 3D digital models of the terminal intra-valley portion up to the outlet onto the Po Plain of the Taro and Trebbia rivers in which are also shown 11 ecological monitoring sites (black dots) located along the two river tracts.(Lower) detailed view from Google Earth of the two investigated tracts.The location of the two main withdrawal channels along the Taro (WW1) and Trebbia (WW2) rivers is also shown.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 3 List of the Sentinel 2 MSI (MultiSpectral Instrument) multispectral optical imagery selected for the multitemporal analysis carried out in this work.

F
I G U R E 6 Multitemporal analysis of the Infrared band 8a for the Taro River (July-August-September 2016) and the Trebbia River (April-June-August 2018) showing as some portions of the two rivers are more affected by changes (Taro: light blue rectangles; Trebbia: pink rectangles) than others.These portions correspond to those reaches of the rivers characterized by wandering channels.[Color figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 7 Multitemporal sequence of images in false colors composition (Infrared bands) over the spring-autumn time span.The pictures show the tendency of wandering channels to thin and disappear especially in 2017 (a) while it is almost not found in 2020 (b), both for the Taro and Trebbia rivers.[Color figure can be viewed at wileyonlinelibrary.com]river infrastructures were found to maintain fixed water depth regardless of lower discharge.In these systems, rivers were converted into a sequence of basins characterized by longer water turnover during low discharge periods but ensuring fish survival.In the case of the Taro and Trebbia rivers, such infrastructures are missing in the studied reaches, and the reduction of discharge results in a proportional decrease of habitat for aquatic organisms and processes.In addition to what is found for rivers over the spring-autumn time span, there are changes that also involve the nearby alluvial areas over the same period.The adjacent areas are portions of the floodplains destined for agricultural use, as shown also by satellite images in which they correspond to the numerous regular-shaped polygons recognizable adjacent to the river.Changes occurring in these areas are highlighted on the FCC scenes by variations of the bluish and yellowish polygons ratio, where bluish polygons correspond to vegetated areas while yellowish polygons to bare soils.As documented by the FCC images, the numbers of areas with bare soils with respect to vegetated areas usually increase during the summer season (i.e., July and August) in the Taro River floodplain while at the end of the summer and the beginning of autumn (i.e., August, September, and October) in the Trebbia River floodplain (Figure9), this trend can generally be recognized over the entire 5-year observation interval as the result of the rhythms that mark the human activity from spring to autumn in the rural areas.
Percentage distribution of the four classes VT (vegetation), BS (bare soil), BS+VT, and WT (water) resulted from the satellite images classification, based on the "Maximum Likelihood" algorithm, carried out in the adjacent plain areas with respect to the Taro (upper) and Trebbia (lower) rivers.As shown by the histograms for 2017, 2018, and 2020, the numbers of areas with bare soils with respect to vegetated areas usually increase during the summer season (August) in the case of the Taro River, while during late summer to early autumn (i.e., August-October) in the case of the Trebbia River.[Colorfigure can be viewed at wileyonlinelibrary.com] removal of such alien invasives as the European barbel Barbus barbus and the top mouth gudgeon (stone moroko) Pseudorasbora parva.For what concerns native species, they belong to rheophilic Cyprinidae (Barbus plebejus, Gobio benacensis, Phoxinus lumaireul, Protochondrostoma genei, Squalius squalus, and Telestes muticellus), Cobitidae (Cobitis taenia) and Gobiidae (Padogobius martensii).It must be remarked that the spawning period and larval development of all these species, most of which are mentioned in the Habitat Directive, fall within the dry season and are strictly influenced by the riverbed morphology.The fish monitoring carried out during 2020 in the Taro and Trebbia rivers has demonstrated a strict correlation between river depth, pool formation, and channel availability with fish size and population structure (work in progress).More precisely, the fish distribution was preferentially restricted to running waters deeper than 50 cm where temperature and hydrodynamism fitted with the ecological requirements of the more abundant cyprinid species.For this reason, fish biomass fluctuated according to the different seasons and water conditions.As an example, during the month of July, the overall fish demography in an area of Trebbia River was in the range 0.10-0.15individuals/m 2 for structured populations of 8 species living upstream of irrigation channels where the mean river depth was 60 cm (44 54 0 54.2" N, 09 35 0 17 00 E).Downstream of the same withdrawal channels (44 56 0 53.8"N, 09 35 0 44 00 E), the general biodiversity fell to five species (0.05-0.06 ind/m 2 ) with individuals of very limited body size (unstructured populations) confined to a marginal number of water pools (medium water depth of 20 cm calculated over the entire river tract).The lateral irrigation channels have consequently become anthropogenic refuges for fish fauna during drought periods.One last important aspect relates to recent European regulations issued to control the so-called alien invasive species (Regulation 1143/2014/ UE).Better management interventions and eradication results are certainly obtained whenever based on correct planning in terms of hydrological conditions and habitat modifications through the seasons.In this way, water pools or lateral channels can be exploited to remove alien invasive species during their aggregation phase.5.3.3 | Ecological flowMaps of wet areas and identification of depth from satellite image analyses can be matched with river discharge in order to construct regressions between discharge and specific river habitats or the number of connections between the main river channel and lateral environments for a given reach or even the number of hydrological interruptions, creating longitudinal discontinuities.The analogy with the calibration of river level and water discharge in a fixed river tract is evident as it is evident as a much greater amount of information (and applications) can be inferred from discharge/habitat or discharge/lateral connections or discharge/wet surface or discharge/ hydrological intermittency multiple regressions.All latter regressions are much more informative than simple level/discharge relationships as they have multilevel implications for river functioning, biodiversity, and the development of river management strategies.Such regressions might also allow back-calculation of river discharge in various reaches and may represent a tool to control water use by multiple stakeholders in a period of the water crisis and conflicts that should stimulate more efficient water use rather than the increase of water withdrawals.
Bathymetric values retrieved from Sentinel 2 MSI image analysis carried out over the spring-autumn time interval for different years.The mean depth of the water column, both for the Taro (upper) and Trebbia (lower) rivers, tends to increase along the river path during the springtime otherwise reaching the minimum during the summer.[Color figure can be viewed at wileyonlinelibrary.com]