INTRODUCTION. THE PROBLEMS UNDER CONSIDERATION

The interest to the use of remote sensing data in the simulation of water and heat regimes in land areas has formed by the mid-1980s, when, on the one hand, rapid development of land surface models began [7, 39, 71, 132], and, on the other hand, the progress in the development of new optoelectronic and radio instrumentation installed on space vehicles, along with remote measurements of land surface characteristics provided data on the space and time variations in those characteristics [68, 126–128, 134], which could be used to extend the information base of the models [13, 110, 133, 142]. Combining these two approaches into one general direction of research on the processes of moisture and heat transfer prompted the conduct of hydrological–atmospheric field experiments HAPEX-MOBILHY (1986) [14, 124], HAPEX-SAHEL (1992, 1991–1993) [51], FIFE (1987, 1989) [129, 130], KUREKS (1988, 1991) [70, 72], EFEDA (1991–1994) [26, 27], BOREAS (1993–1996) [108, 131], MOPEX (1996–2003, 2004–2009) [15, 42, 123], etc., the organization, under the World Climate Research Program WCRP (since 1980), of a project of global energy and water exchange GEWEX (since 1990) [141], as well as the implementation under international geospheric–biospheric program IGBP (1987–2015) [62], of the main BAHS program (since 1993) [24, 59]. The objective of practically all field experiments mentioned above was to measure the land surface (LS) parameters, including the characteristics of vegetation, as well as water and heat fluxes and the fluxes of matter (as a rule, with the incorporation of the data of earth remote sensing (ERS)), with the aim to evaluate the model parameters of the interaction between LS and the atmosphere and to reveal the role of the biosphere in this interaction. The objective of the GEWEX project is to study the energy and water cycles of the Earth and to evaluate the water and energy fluxes at the regional and global scales [141]. The objective of the IGBP program is the formation of the knowledge on the behavior of various elements of the Earth system under the effect of physical, chemical, and biological processes, as well as the description of the anthropogenic effect on global processes involved in the water cycle as well as the cycles of carbon, nitrogen, sulfur, and phosphorus. The BAHC project was intended to study the effect of biosphere elements, in particular, vegetation, on the dynamics of the hydrological cycle by carrying out experiments and simulating processes of the formation of energy, water, carbon dioxide fluxes, and sediments in the system soil–vegetation–atmosphere for various space and time scales. As the considerable effect of vegetation on the formation of these fluxes makes it an important factor of hydrological cycle regulation and climate formation, an important task of the project was the identification of vegetation cover transformations caused by human activity and considered as a factor of climate changes. The BAHC program also included studying the effect of the climate and human activity on the accumulation and river transport of sediments [59].

The data collected by surface measurements and ERS, as well as by simulating moisture and heat exchange were, to a certain extent, combined to form regional and global databases on LS and weather characteristics, including those extending up to now and newly formed. These include the systems and databases GCOS [50], NCEI NOAA [102]—ISD [63], and LCD [77]. The observation system GCOS, created in 1992 with the support of WCRP, determined 50 key climate variables applicable to global climate observations. Considerable contribution to GCOS was made by the World Integrated Global Observing System WIGOS WMO and the World Hydrological Cycle Observing System WHYCOS with databases. ISD is a global database of hourly and synoptic observations of the characteristics of LS and surface atmosphere layer from numerous (more than 100) sources from more than 20 000 stations. These data are collected in a common format and combined to form a single data model. The LCD base contains sets of climate data, obtained by generalization of local climate conditions for more than 2500 weather stations and airports of the USA. ECOCLIMAP-I database [5, 91] and ECOCLIMAP-II [46], improved with the use of datasets for European territory, are used to initialize (to specify the initial conditions and parameter values) models of SVAT type for meteorological and climate models (at all horizontal scales). At the formation of ECOCLIMAP-I database for regions with homogeneous vegetation, identified by combining LS maps and climatic maps with the incorporation of AVHRR data with the use of fields of LAI values, constructed by these data, the values of all model parameters were constructed (the characteristics of vegetation cover and weather characteristics) [91]. The construction of ECOCLIMAP II database used a more complex specification of the examined areas by LS and plant types (some clusters) and incorporated two proxy variables NDVI and LAI, determined by data of radiometers SPOT/Vegetation and MODIS, respectively. The values of all parameters of model ISBA, class SVAT (developed by Météo France) were obtained for these clusters [46]. The global database ISLSCP II, formed in ORNL DAAC NASA under the project with the same name, a part of GEWEX project, contains archived overall databases of 1986–1995 on the LS characteristics; hydrological, meteorological, radiation, and soil characteristics with resolution of 1/4°, 1/2°, and 1° [29]. These archives were supplemented by sets of estimates of LS–NDVI, albedo, LAI, and other characteristics by satellite data of sensors AVHRR, MODIS, SPOT/Vegetation, and VIIRS over later years [29]. The part “Hydrology and Soils” in this database contains, for example, sets of estimates of the proportion of soil moisture available for plants, estimates of the monthly runoff volumes, datasets on daily, pentad, and monthly total precipitation, as well as data on precipitation, required for solving climate problems [64], as well as many other datasets.

In the recent 25 years, global and continental GIS-databases in different fields were developed including hydrological, climate, and soil ones. The global databases include, for example, USGS database LCI on land cover; IRI/LDEO CDL database of the Earth Institute, the University of Columbia and the Observatory of Land Lamont-Doherty at the same university, containing datasets from more than 300 climate models and databases; the database of soil characteristics HWSD FAO [45], containing data on the texture parameters, depths, and soil acidity; the global elevation database ASTER GDEM with a 30-m resolution, constructed based on satellite images of ASTER radiometer; a set of renewable maps of soil properties and classes in the world with a resolution of 1 km (SoilGrids1km—soil property and class maps), compiled with the use of modern statistical methods; the global database CHELSA with a resolution of 1 km, the first implementation of which [66] contained sets of air temperature estimates, formed at a statistical decrease of the scale, and monthly precipitation totals, obtained by data of climatic reanalysis ERA-Interim with the use of wind speed, valley slope aspects, and the heights of the atmospheric boundary layer. The second implementation [30] included sets of monthly estimates of air pressure deficit, incoming short-wave radiation, potential evapotranspiration, climatic water humidity index, and area water index over 1980–2018. These estimates were formed based on the results of calculation with combined data obtained by joining the data derived by a decrease by delta-method of the scale of time series of relative air humidity near land surface and the proportion of cloud area along with a mechanical decrease of the scale for the data on temperature, precipitation, and solar radiation. The GIS-databases of global level also include OpenAerialMap with an open license, which provides a set of instruments for the search, exchange, and use of images from satellites and UAV (Unmanned Aerial Vehicle) and OpenLandMap, containing sets of various data on the Earth (the so-called Earth mask) on LS, vegetation, soils, climate, information on terrain etc.

One of the continental GIS bases is the European base ESDAC [112], containing information of the European Soil Data Base, sets of estimates of soil erosion, the amount of organic carbon in soil, its biodiversity, as well as data of LUCAS (results of laboratory physical and chemical analyses of the top (0–20 cm) soil layer for samples taken at more than 20 000 sites in the European Union countries), etc. We will also mention two GIS-databases for USA: NGDC—a base of digital elevation models (DEV) in free access, data on LS, seismological and other data, as well as a soil database of US Ministry of Agriculture (USDA NRCS). The latter can be combined with various datasets with the use of an online mapping instrument Web Soil Survey.

In the recent decades, considerable attention was also paid to the formation of ERS databases on LS and weather characteristics for their use in simulating the processes of moisture and heat exchange. For example, the sources of such data in open access include: the database USGS Earth Explorer for various territories, containing Landsat data, unclassified 1960–1970 data from satellites of CORONA system, geospectral data, obtained with the use of NASA Hyperion spectrometer; Sentinel Open Access Hub with survey data of twelve-channel multispectral camera in the ranges from the visible to SWIR from Sentinel-2 board at 10-m resolution in VNIR ranges and the survey data with the use of SAR in the microwave C-range from the board of Sentinel-1 with a resolution from 5 × 5 to 20 × 40 m at the quality of images poorer than that of Sentinel-2; NASA Earthdata Search base (the majority of data enters into it from the NASA DAAC center), representing products for the analysis of Earth processes, is a high-class source of data on the global land use and vegetation cover and contains specialized satellite data, such as the types of permafrost and water–swamp fields; the set of NOAA Data Access Viewer, containing satellite images, aerial photographs, and images obtained with the use of lidars; database of DigitalGlobe—a commercial operator of high-resolution satellites GeoEye-1, QuickBird, IKONOS, WorldView group, from which the entire library of 30-cm images DigitalGlobe can be loaded free of charge. In the case of any natural disaster, the Open Data Programme DigitalGlobe for help presents, free of charge, satellite images of fires, floods, hurricanes, typhoons, and earthquakes. Also, data samples can be obtained free of charge, including contours of buildings, elevations, stereo images, and images in real color. The database of the Geo-Airbus Defense company, a commercial supplier of satellite images from SPOT, Pleiades, and RapidEye satellites, includes a set of image samples; however, the choice of satellite images free of charge is very limited. The proposed samples include optical images SPOT with a resolution of ≤1.5 m, radar images TerraSAR-X with a resolution of ≤3 m and horizontal sections of DEM WorldDEM with a resolution of 12 m (more accurate than DEMs ASTER, and SRTM). The database of the National Institute of Space Studies of Brasilia (INPE) is a catalogue of INPE images, similar to the library for free of charge loading of satellite images. The major portion of this catalogue consists of images from CBERS Chinese–Brazilian satellites, and all the data refer to the South America and Africa. It will also contain the data from CBERS-4 [114], Indian satellite ResourceSat and British UK-DMC 2. The global DEM AW3D30 with a 30-m resolution is an open-access set of world level data based on the results of surveying from onboard satellite ALOS JAXA by PRISM instruments in the optical range with a resolution of 2.5 m. This set was formed based on a dataset from DEM version with a 5-m grid World 3D Topographic Data, which now contains the most accurate data on elevations at a global scale. The web-site VITO Vision presents free of charge images of satellites PROBA-V (PROBA-Vegetation), SPOT-Vegetation, and MetOp. These satellites with a low-resolution cut patterns of vegetation on the Earth surface. The Satellite Land Cover database at a global level gives the sources of data from Landsat, MODIS, and AVHRR on the land cover, which allow observing quantitative changes in its characteristics all over the world—for territories with various geological and hydrological conditions, various vegetation and agricultural features, as well as for urban areas. An extending database, used in the simulation of moisture and heat exchange, is the GLASS base, the number of satellite products in which increased from 5 in 2012 [84] to 14 in 2020 [83]. Now, the database contains datasets on LAI, broad-band albedo A and longwave radiation capacity E, incoming shortwave R and resultant longwave radiation, photosynthetic active radiation PAR and its adsorbed fraction FAPAR, LS temperature LST, projective cover by vegetation B, evapotranspiration ET, latent heat flux LE, etc. [83]. The duration of uninterrupted measurements of LAI, A, E, R, and PAR exceeds 35 years (since 1981 until now) without gaps, the space and time resolution of the former three is 1–5 km and 3 h, respectively. The Landsat USGS database [74], which has been updated with the results of sounding from Landsat-9 since February 2022, is also widely used [75]. The global data of Landsat-9 OLI-2 in nine ranges (VNIR, SWIR, and others) with a resolution of 30 m and TIRS-2 in two thermal ranges [76] enable numerous estimates of the characteristics of PP and atmospheric boundary layer, as required for the implementation of hydrological models. The use of data on land classification, carried out in accordance with IGBP classification scheme, from the dataset of terrestrial ecosystem classification MODIS MOD12Q1 [98], as well as daily, eight-day, and monthly data on the albedo and NDVI, obtained by MODIS measurements with a resolution of 1 angular minute [97], combined with Landsat-9 data, will improve the accuracy of calculations using these models.

The launch in the recent 15–20 years of the satellites mentioned above with instruments of either high or low resolution has resulted in the appearance of additional data that can be used in models for areas of various scales, from local to global, and, therefore, the formation of appropriate databases. These are GLCC [87], GLDAS [120], NLDAS-1 [96], and NLDAS-2 [155] and other LDAS [154]. The launch into orbit of satellites MetOp-A, -B, -C, SMOS, SMAP, GCOM-W1, and GCOM-C1, as well as above-mentioned Sentinel-1, and others with active or passive microwave range sensors, as well as with radars, allowed the obtained data to be used to calculate soil moisture content in areas with a size determined by satellite equipment resolution and varying from an individual field to a continent [44, 69, 107, 116, 118, 137, 138, 145]. The satellite imagery with the use of synthetic aperture radars, which has been actively developing in recent years, attracts the interest of users because of its all-weather capability, i.e., the independence of cloudiness, which is typical of data of VNIR ranges, and the ability of radio waves emitted from a satellite to penetrate under tree crowns and next to be recorded in the form of a reflected signal, thus presenting data without time delays. Examples of the implementation of this technology include 1-st generation satellites (2007–2012) TerraSAR-X, Radarsat-2, COSMO-SkyMed, TanDEM-X and 2-nd generation (since 2018): SGS-1, PAZ. Note also that the Sentinel-1 images with 10-m resolution with resurvey within 6–12 days were found to be acceptable for the established community of SAR data users. Now these data are used to study Antarctic icebergs and to reveal environmental changes, as well as to map the effects of natural or anthropogenic impacts. Examples of such use are tracing the effect of explosions, destruction, forest fires, and flood scales; monitoring marine and coastal zones, including oil spills at tanker accidents; monitoring of land, in particular, determining the location of land subsidence and domains subject to the risk of collapses and landslides; studying the state of vegetation (the formation of canopy, the density and growth rates of crops) and the specific features of agricultural activities (the determination of field boundaries and sizes, as well as the time of planting and harvesting). At the passage in the recent 5–7 years to the formation of constellations of small-size microsatellites with a weight of <100 kg, due to the much lower cost of their launch compared with the conventional satellites, the potential of the use of SAR data have increased dramatically. Finnish microsatellite ICEYE-X1 appeared in 2018, followed by 16 other satellites of ICEYE company; satellites SAR Denali (2018) and Sequoia (2020) were launched by Capella Space (USA), followed by 5 global level satellites with hourly time resolution. The maximal number of CubeSat microsatellites (almost 200) was launched by the Planet company. Their data are actively used to solve the problems described above.

In addition to the incorporation of data from various carriers in the simulation—conventional satellites with equipment of various spectral bands, microsatellites with VNIR sensors and radars, aircrafts, and UAV, these data are often combined with models to form appropriate databases, e.g., GLDAS Noah database—a symbiose of the Global Land Data Assimilation System and LSM Noah. Such combination contains a larger volume of data than the results of common observations [92, 143]. The sets of simulation results thus formed are compared with the results of measurements collected in databases created under programs NASA GRACE [173] and its continuation (since 2018) GRACE-FO [52] which are among the most informative. They are supplemented, in addition to the conventional data, by the data of ICESat-2 satellite [36, 60, 61] and will be supplemented by the data from WCOM [135–137] and SWOT [23, 43, 54] scheduled for launch. In the implementation of the programs analogous to GRACE-FO, the appropriate databases of terrestrial and satellite data are also extended, along with the data of reanalysis and the results of simulating the processes of moisture and heat exchange, e.g., OSCAR/Surface and OSCAR/Space [109], working in the WIGOS system [149, 150].

The development perspectives of satellite technologies for evaluating the characteristics of land surface and the surface layer of the atmosphere and the use of the obtained estimates for simulating moisture and heat exchange include the construction of new high-resolution multispectral instrumentation, the formation of vast databases, including cloud, and methods for their processing and assimilating in models. The acquisition of such data with the use of numerous microsatellites with low launch costs and with sensors in various ranges (from VNIR to microwave), showing a high (1.5–4 m) resolution, and covering land surface several times a day, is also highly demanded and promising.

Within the range of problems described above, this review will consider the issues of satellite data utilization in simulating elements of water and thermal regimes of land areas, i.e., the vertical fluxes of moisture and heat, and soil moisture content.

USING REMOTE SENSING DATA IN SIMULATING ELEMENTS OF WATER AND HEAT REGIMES OF LAND AREAS

The Use of Satellite Data in Simulating Vertical Fluxes of Moisture and Heat

Simulating the water and heat regimes of various territories implies reproducing the dynamics of soil water content, evapotranspiration ET (i.e., the total evaporation, including evaporation from bare soil and transpiration by plants), the vertical heat fluxes LE and H, as well as other components of the hydrological cycle, commonly, with the use of models of water and heat exchange between the land surface and the atmosphere, for example, of LSM class, containing equations of water transfer and heat conduction [9, 111, 115, 140]), or balance models of the type of SEBAL, METRIC, SEBS, TSEB, etc. [20, 21, 80, 88]. The information basis of LSM is the sets of soil, vegetation, and snow cover characteristics, used as input variables [8, 9, 93, 101, 140]. Under conventional approaches, these values were derived from ground observations, while now they are mostly taken from Earth remote sensing. These data are used in LSM to evaluate parameters, e.g., the characteristics of soils and LS; to reveal precipitation zones and amounts; to specify the initial conditions for the model, e.g., soil moisture SM, and to construct time-varying estimates of LS state, e.g., snow water equivalent SWE. The set of hydrological characteristics used in LSM and the methods for their assimilation in models are given in [55]. Estimates of some of these characteristics, i.e., vegetation index NDVI, LS emissivity E, vegetation cover fraction B, leaf area index LAI, land surface temperature LST, and precipitation, were constructed with the use of the developed procedures for thematic processing of measurement data in the visible and IR channels of radiometers AVHRR/NOAA, VIIRS/SNPP, MODIS/Terra, and Aqua, MSU-MR/Meteor-M, SEVIRI/Meteosat [1–4, 10, 11]. In [93], the values of albedo, NDVI, and LAI were determined using MODIS data. Many studies in this field, carried out with the use of balance models were aimed to obtain estimates only of ET or LE [12, 22, 32, 144, 147, 160]. In such studies, the values of these variables were determined by different methods: calculated by Penman-Monteith [80, 88, 144] or Priestley−Taylor [32], evaluated with the use of lysimeter measurement data [22, 144], or as a residual term in the radiation balance equation [12, 22, 82, 144, 147], by measurements of flows from towers [80, 88, 153] or scintillometers with a large aperture LAS [80], or estimated based on data of MODIS/TERRA (product MOD16A2) and MODIS/AQUA (MYD16A2) [80, 144]. In [16], the LE (and ET) and H fluxes were calculated using the SEBAL model, in whose equations the albedo and LST values were introduced, which were determined using data from the OLI and TIRS sensors of the Landsat-8 and MODIS/Terra satellites (the product MOD09A1). The error in ET estimates did not exceed 0.35 mm/day and lied between 11 and 12.5%, and that for H estimates lied between 26 and 35%. The study [103] gives estimates of subsoil water level at the construction of its regression dependences on the potential evapotranspiration, calculated with the use of a balance algorithm of SEBAL model. A common feature of all these models is the use of data on vegetation characteristics—NDVI, E, and LAI, obtained from data of MODIS [32, 144, 147], VIIRS [80], Landsat-5, -8 [22, 103, 147], as well as data on heat fluxes, which are used to calculate LST and LE, determined by data of MODIS and Landsat [32, 144, 147], Landsat-5, -8 [22, 49, 103] with a resolution of 30–120 m, MODIS and VIIRS [80], and ASTER [88]. Measurements of airborne radiometers are also used to obtain estimates of NDVI, B, and LAI [153], as well as heat fluxes [49]. The latter estimates are compared with the results of measurements from “flux towers” [153]. Estimation of the vertical fluxes of moisture and heat at different levels above the land surface were also carried out in the framework of a complex experiment LAFE [152], in which Doppler lidars were used to determine wind speed, air temperature and absolute humidity; airplane and helicopter data on an area 10 × 10 km were used to determine the albedo, NDVI, LAI, and LST; soil moisture was determined by network data; radiation fluxes were measured by vortex covariation method on flux towers installed on SEB-stations. The time resolution of lidar data was 1–10 s, and their space resolution was 30–300 m. In many studies, a part of source data was taken from satellite and surface databases; thus, in study [12], a set of plant characteristics was loaded from archives SPOT-Vegetation data and Global Land Cover Map, and estimates of LST and radiation fluxes, from LSA SAF database with a 15-min time resolution and 3-km space resolution.

A similar approach was implemented at a global level in studying the agreement between estimates of evapotranspiration, obtained with the use of satellite data by three different methods and input data sets, including CSIRO-PML [163], MOD16(A2) [99], and GLEAM [86, 89, 90, 94, 95], as well as estimates of precipitation based on GPCP data [57] and changes in soil water content by GRACE data [92] for the basins of Lake Erie, the Aral Sea, and the rivers of Colorado and Niger [86]. The spatial resolution was 0.5° for MOD16 (A2), 0.25° for СSIRO-PML and GLEAM, 1° for precipitation GPCP (1DD v1.2) [58], and 333 km for GRACE. Estimates of vegetation cover characteristics (NDVI, LAI, etc.), the data on LST, radiation fluxes, weather data, including precipitation, were taken from various bases and archives. The increase in the last 3–4 years of the amount of information used to estimate the vertical moisture and heat fluxes increases the possibilities for implementing modern LSM such as SURFEX, JULES, ORCHIDEE, CLM5, and CABLE, for territories at various space scales with the use of mosaic subgrid structure, when using a mosaic subgrid structure, which makes it possible to disaggregate sets of required estimates of water and energy characteristics from satellite and ground-based databases [25, 48].

The Use of ERS Data for Evaluating Soil Moisture Content

The need for estimates of soil water content W for territories of different size is determined by the significance of this characteristic as a component of their water balance—an indicator of the state of water resources. For example, K. Zhang et al. [162] evaluated changes in subsurface water content for 168 river catchments as an overall response to global changes in the values of precipitation, evapotranspiration, and runoff volumes. The data on subsurface water content were taken from GRACE information; the estimates of soil and vegetation water content, water content on land surface, as well as snow water equivalent were compiled from GLDAS databases; two datasets on precipitation were taken from CRU and GPCC databases. Estimates of evapotranspiration were formed using MODIS (MOD16A2) data and taken from GLDAS database, the values of potential evapotranspiration were taken from CPU dataset; and the data on runoff, from study [37]. A digital map for 168 catchments was obtained from GRDC. The contribution of precipitation to the global changes of subsurface water content, determined by GRACE data was 42.6%, that of evapotranspiration was 43.2%, and that of runoff was 14.2%. The obtained results can illustrate the possible structure of such changes under changing climate. The water content W are also an important characteristic of available water supply, they determine the dynamics of plant growth and, as a consequence, agricultural production, which can be seen in the case of deficiency of such reserves and droughts. Estimates of W are also required for calculating runoff characteristics and forecasting floods and freshets, especially, in large rivers.

Soil Surface Moisture Estimates and Water Content of Its Active Layer Using Microwave Remote Sensing Data

In the recent 20 years, the number of estimates of W based on ERS data of the microwave range and obtained at any weather has increased considerably. Depending on the problem to be solved, these estimates were carried out for territories with areas ranging from a few to hundreds of thousands of square kilometers with the use of the measurement results with radiometers AMSR-E/Aqua [111] and AMSR-2/GCOM-W1 [6, 107, 138] in C-, X-, K-, Ka-, and W-ranges with a resolution for AMSR-2 from 35 × 62 km at a frequency of 6.9 HGz to 3 × 5 km at a frequency of 89.0 GHz [6, 107]; radiometer MIRAS of SMOS satellite in L-range with a mean resolution of 43 km [69]; radiometer and radar of SMAP satellite in the same range with a resolution of radiometer of 40 and radar of 3 km [44, 118]; scatterometer ASCAT/MetOp-A, -B, -C in C-range with a resolution of 12.5 km [19, 145]; radar of satellites Sentinel-1A, -1B within the same range with a resolution from 5 × 5 to 25 × 100 m [17, 18]. Since the spatial resolution is proportional to the antenna diameter and inversely proportional to the wavelength, at small wavelengths, an antenna of large diameter is required to obtain high resolution at sounding from a height of several hundred kilometers, which causes considerable engineering problems [69]. Note that, to solve some practical problems under H-SAF [56], a procedure was proposed for disaggregation of the estimates of soil surface humidity (SSH) using ASCAT data with a resolution from 25 to 1 km, based on the use of a linear regression equation with the incorporation of the ASAR radar data from ENVISAT satellite and in-situ data [145]. At the described scatter in the microwave equipment resolution, the values of W can be estimated at a global and regional scale, at a level of individual river catchment, and even a field. L. Brocca et al. [28] used data of an AMSR-E/Aqua radiometer (with the use of three different algorithms) and an ASCAT/MetOp scatterometer to evaluate SSH down to the depth of 5 cm in 17 areas in Italy, Spain, France, and Luxemburg. An exponential filter was used to calculate soil moisture SM in the root zone. The correlation between the obtained estimates and observation data was largest for the results based on ASCAT data for all areas in France and central Italy; the results for other regions are identical. It was also found that an increase in vegetation density leads to a decrease in the reliability of all satellite estimates of SM. The estimates of SM formed for various applications with the use of measurement data from microwave range sensors, such as AMSR-E, TRMM-TMI, SSM/I, WindSat, ERS-1, and -2, ASCAT, as well as SMOS and, later, SMAP, obtained with the use of LSM or reanalysis results, as well as from prognostic centers, e.g. ECMWF or GLDAS and GSWP databases, were combined into ISMN in 2011 [41, 65]. The calibration and verification of the satellite and model estimates of soil moisture were made using in-situ measurement data with a 1-h interval from reliable sources. These data provided information on the space and time variations of SM at different scales. W.A. Dorigo et al. [41] give a review of direct and indirect methods of in-situ SM measurements (gravimetric, with the use of neutron humidimeter and a cosmic ray neutron detector, by the difference between the dielectric constants of soil components and water, with the use of tensiometers, by changes in soil temperature associated with its water content, as well as other methods). As of May 2011, ISMN contained data of 16 regional networks and more than 500 stations in Italy, France, the United States, Russia, Mongolia, Australia, France, West African countries, and others for a period from 1952 to 2011, and this database was expanded in the following years.

J. Shi et al [137] used airborne radiometer measurements of the L- and S-, as well as L- and С-ranges for various space scales to construct SM profiles from 0.1 to 0.4 m in depth, which were compared with measurement results. In this case, the main interfering factors were surface roughness and the presence of vegetation. These results showed that the joint use of L‑band and S-band devices improve the accuracy of SM estimation due to the better correction of the effects of vegetation. In addition, the data of observations in the S- and C-bands have a higher spatial resolution than that of data in the L-band; therefore, they can be used to reduce the scale of L-band observation data from, for example, SMOS and SMAP and later WCOM satellites. Since the microwave band data are used to directly estimate the surface (0–3 cm) soil layer, the water content of its active layer are evaluated with the use of several approaches. The first approach consists in simulating the process of water transport into deeper soil layers with the use of SWI time series, which have an exponential autocorrelation function with a characteristic time that is consistent with the theoretical expectation time and observation results [145]. The second approach is based on the application of extended Kalman filter, which takes into account the errors of the prognostic model without limitations on the time interval of acquisition of earlier satellite data with the division of the soil mass into four layers: 0–7, 7–28, 28–100, and 100–289 cm [33, 34, 73, 120, 145]. The studies [116, 118] use Kalman ensemble filters, in which their ensembles are used to estimate the covariation matrix of forecast errors. The source information used in the studies are ASCAT data in [33, 34, 73, 120, 145], SMAP data in [118], and Sentinel-1 data in [116]. Under the third approach, W is estimated with the use of LSM. In SVAT models, the vertical profiles of soil moisture content are calculated with the use of water transfer and heat equations [8, 9]. The estimates of the characteristics of vegetation, precipitation and LST are constructed using measurement data from meteorological satellites. In the calculation of water content, daily SSH composites, formed based on ASCAT/MetOp data, were used at each time step to specify the initial profiles of SM as well as to evaluate evaporation from soil surface with the use of bulk-formulas at each time step [8, 101]. Examples given in Fig. 1 show time variations of SSH over the vegetation season, constructed for a station in Volgograd region based on ASCAT data and by simulation results, and Fig. 2 shows the distribution of soil water content in a part of the Central Chernozem Region of Russia (TsChR) for a day in the vegetation season of 2017, calculated by the model with the use of SSH estimates based on ASCAT data and by surface data [101].

Fig. 1.
figure 1

Soil surface humidity calculated with the use of the model (1) by data of ground measurements and (2) with the use of evaporation estimates obtained with the use of ASCAT/MetOp data, and (3) determined directly by ASCAT measurements. Danilovka Agrometeorologial Station (Volgograd province). 2019 Vegetation season.

Full size image

LSM Noah [35, 105] is also used to evaluate values of W (at both global and regional scales). In [161], such estimates obtained with the use of this model, associated with GLDAS database, and with the use of ECV, European reanalysis ERA-Interim, and MERRA data with a spatial resolution of 0.25° × 0.25°, 0.5° × 0.5°, and 1° × 1°, were compared with the results of 100 in situ observations for three areas of Himalaya–Tibet plateau. In this case, Noah model generally showed better results. We note the difference between the used data sets. GLDAS-1 and GLDAS-2 are databases of estimates of LS and fluxes, which were formed by joining the results of satellite and ground-based measurements with the use of LSM and data assimilation methods. In ECV, the sets of W estimates are combined from the measurement data of various sensors from radiometer SMMR/Nimbus-7 to scatterometer ASCAT/MetOp. In ERA-Interim, the results of reanalysis were obtained at an alternative regime of data assimilation with the use of 6-hour cycles of analysis, and contain estimates of W for four soil layers (0–7, 7–28, 28–100, and 100–289 cm) [158]. MERRA is a set of reanalysis assimilation results, containing ERS and ground-based observations of the atmospheric characteristics, as well as radiation data from probes, and wind data from scatterometers. LSM Noah GLDAS along with other three models—BHOA [47], ERA-Interim TESSEL, and LISFLOOD [31]—and satellite data ESA CCI was used to evaluate variations of soil surface moisture and water content of its root layer with the aim to compare them with in-situ data from eight sites in Argentinian pampa [139]. It was found that the time distributions of soil water content constructed based on the data of model and satellite estimates are incorrect in some cases, and to extend the possibility for detecting droughts or overmoistening with the use of LSM, it is necessary to have more accurate weather data and the values of soil characteristics.

Miralles et al. [94] calculated the soil water content W in the root zone for global scale with the use of GLEAM model with balance equations solved by difference scheme for successive soil levels. The input variables were the meltwater volume and the total rain precipitation minus the volume of moisture intercepted by plants, and the output variables were the values of evaporation and moisture infiltration into deeper soil layers. The estimates of water balance were corrected for each day with the use of the Kalman filter for satellite observations. The evaporation was calculated by the Priestley–Taylor formula for bare soil, low (0–5, 5–100 cm), and high (0–5, 5–100 cm and 1.0–2.5 m) vegetation at the use of estimates of SSH, LST, plant cover density, and snow depth, obtained using measurement data from AMSR-E/AQUA radiometer in the 36.5 GHz band with a resolution of 12 km and within the 6.9 GHz with a resolution of 56 km. Precipitation was determined by observations with a half an hour interval within IR-range by geostationary satellites GOES, GMS, and Meteosat for increasing the volume of their more accurate estimates in the microwave range by data of sensors AMSU-B, SSM/I, TMI, and AMSR. The set of data on vegetation VCF product was formed using MODIS data (MOD44B package). Air pressure was calculated by the barometric height formula with the use of digital elevation model. Evaporation estimates were verified using data of FLUXNET stations in Europe and USA for the most typical climate conditions and vegetation types. Such approach based on the use of GLEAM to evaluate the water content W of the root layer and evaporation ET, was also used in [95]. Both values were calculated with the use of water balance equation with the incorporation of data on runoff from the basins of 24 large rivers; the evaporation was calculated by Priestley−Taylor formula, and the water content were estimated with the use of Kalman filter. The correctness of the daily ET estimates for the global scale was checked by their comparison with the appropriate values from Princeton University product and MERRA databases. The former were obtained with the use of Penman–Monteith equation with incorporation of data from various satellites. The latter were based on the reanalysis data from GEOS-5 DAS NASA. The results of calculations for different months were used to construct maps of evaporation and transpiration for the globe. Analogous estimates of soil water content W in the root zone and evaporation, including a set of calculation results over 36 years (1980–2015) (v3a), were obtained in [90] with the use of a new (v3) version of GLEAM, containing the same input variables as the previous (v2) version of the model [94], also with the incorporation of satellite data. The version v3 used Priestley–Taylor formula, Kalman-filter algorithm, MOD44B package, and other products. For low vegetation, the root zone was considered in a two-layer variant (0–10, 10–100 cm), and for high vegetation, in a three-layer variant (0–10, 10–100, 100–250 cm). At a three-layer description, the water content were calculated with the use of finite-difference schemes for each horizon. New estimates of W and ET were obtained when modeling based on the data of measurements of various passive and active microwave sensors in C- and L-ranges of ESA CCI (for the set of results over 2003–2015, v3b) and data of SMOS satellite (for an analogous set over 2011–2015, v3c). The adequacy of the estimates of W and ET from sets v3а, v3b, and v3c is confirmed by their comparison with the measurement results of 2325 soil moisture sensors and fluxes at 91 flux towers by vortex covariation method for various types of LS. The comparison of W estimates from data sets showed their quality to be higher than that of analogous estimates from v2. All-weather satellite data in microwave range with a high resolution (100 m) were used in GLEAM to evaluate the ET and SSH for the Netherlands, Flanders, and West Germany [89]. SSH and LST were evaluated at a lesser scale with the use of LPRM model using measurements of AMSR-2 radiometer on GCOM-W1 satellite. The surface moisture content was also determined by the data of scatterometer ASCAT/Metop-A, -B. The water content in the root zone W were calculated with the use of a multilayer water balance model—a part of GLEAM, based on the observations of precipitation with assimilation of data on SSH. The datasets that enter the used version of GLEAM included estimates of potential ET from water surface, low and high vegetation, and bare soil with the use of Penman–Monteith equations and Priestley–Taylor constants. The data on short- and long-wave radiation were obtained from measurements of radiometer SEVIRI/Meteosat with resolution of 5 km, as well as from EUMETSAT LSA-SAF archive; the data on albedo were taken from measurements of MODIS/Terra and Aqua with 500-m resolution once in 16 days (MCD43A3 package). SSH estimates were compared with in-situ measurements at 29 sites in the examined territories, the mean coefficient of correlation was 0.76. The comparison of ET estimates obtained with the use of GLEAM with LSA SAF data revealed difference between them (coefficients of correlation of 0.65–0.95) depending on the vegetation type—for woody vegetation, the errors were greater than those for grassy and shrubby vegetation because of the interception of water by plants, which was the dominant process. Combining the data obtained from passive (AMSR-E) and active (ASCAT) microwave satellite sensors and the use of a developed technology [85] gave more accurate estimates of SSH for a soil layer with a depth up to 10 cm on a global scale. Such and analogous estimates, obtained with the use of Noah model (a component of GLDAS), were compared with the results of in-situ SSH measurements, taken from OZNET networks for the southeastern part of Australia, REMEDHUS for the central Spain, SMOSMANIA for the southern France, and CNR-IRP for Italy. These comparisons show that at a good correlation between both satellite products (R > 0.65), their combining improves the time resolution of SSH estimates. In addition, the use of scaled estimates AMSR-E and ASCAT was found to be effective for regions with sparse or moderate vegetation. K. Zhang et al. [159] showed that the accuracy of the daily estimate of soil water content based on the joint use of observation data on the brightness temperature from several satellites in the regime close to real time is higher than the accuracy of analogous estimates by data of each satellite. Estimates of soil moisture were based on the measurements of the brightness temperature by SMAP, SMOS, AMSR-2, FY3B, and FY3C equipment for the same day, followed by averaging these estimates to obtain their daily aggregate. The comparison of the combined estimates and those from each satellite with ground-based data from two networks in the Central Tibet and Anhui province (China) showed a noticeable increase in the accuracy of soil moisture estimates, especially in the second region. B. Zhu et al. [167] present a simplified algorithm for determining the moisture of bare soil surface with the use of measurements of L-range radiometers installed on SMOS and SMAP satellites. This algorithm consists of two parts: a model of surface radiation with dual polarization, which allows the effect of surface roughness to be reduced by using data on the reflectivity of the surface, and a model of soil moisture recovery, which utilizes the relationship between the corrected actual refractive index and the volumetric soil moisture content. In this approach, SM was determined with the incorporation of several available input characteristics: the brightness temperature of the microwave range at dual polarization, the surface temperature, and the content of sand and clay in the surface (0–10 cm) soil layer. The obtained estimates showed good agreement with the SM values calculated with the use of an appropriate integral equation (the RMSE values were <3% at all angles of incidence). The subsequent test of the algorithm with data of four-year experiments with radiation in L-range, carried out in BARC, yielded RMSE of 4.3 and 3.4% at the angles of incidence of 40° and 50°, respectively. The algorithm also showed a high efficiency for radiometers in L-range with larger angles of incidence installed on SMAP.

Estimation of Soil Surface Moisture Using Neural Networks

SSH estimates by satellite data can be also obtained with the use of neural networks NN [100, 121, 156]. N.J. Rodríguez-Fernández et al. [121] choose this set of estimates by SMOS data as a reference in learning NN to obtain a global nonlinear dependence (search for global nonlinear regression), relating the brightness temperatures Tbr, measured by AMSR-E radiometer, with SMOS data set over the period of simultaneous mission of 1.5 year, after which the constructed network was used to evaluate SSH by the earlier AMSR-E observations. The sensitivity of Tbr AMSR-E to changes in soil temperature was evaluated with the use of information from databases ERA-Land and MERRA-Land. One of the articles, where data of C- and X-ranges AMSR-E and information about soils was used, was chosen to obtain SSH estimates over 2003–2011. This set of estimates showed small biases (<0.02 m3/m3) and standard deviations (<0.04 m3/m3) from SMOS data over most of the world surface. SSH estimates from this set, as well as from sets AMSR-E, SM CCI, MERRA-Land, and ERA-Land were compared with many in-situ measurement results on four continents. A good agreement with field data was obtained for Australia, the results of calculations were found acceptable for Europe, mountain regions of the North America, and, to a much lesser extent, to the Sahel region. They also demonstrated considerable errors for regions of tropic and boreal forests. A similar approach was implemented in [156], where backpropagation neural networks BPNN were used to reconstruct a long time series of SSH values with the use of satellite-based estimates of the microwave vegetation index MVI. Here, BPNN training for each grid pixel was carried out for two years with the use of data on SSH from SMOS of level 3 (SMOSL3sm) as a training objective, taking into account the reflection indices Rs within ranges C/X/Ku/Ka /Q, as well as MVI from the data of radiometers AMSR-E and AMSR-2 as input characteristics, where MVI was used for correction for the effect of vegetation. In this case, the values of Rs and MVI were determined from the relationships with the brightness temperature Tbr, measured by AMSR-E and AMSR-2. The training accuracy of the networks was evaluated by comparing the SSH values obtained with the use of BPNN (NN sm) with SMOSL3sm within the training period BPNN, by correlation coefficients (~0.67), bias (~–0.0005 m3/m3) and the root-mean-square error RMSE (~0.055 m3/m3). Good results at a global scale were obtained for Australia, the central part of USA, and Central Asia. The trained networks were used to construct for each pixel, according to AMSR-E data, global SSH time series over period 2003–2015, after which the products NN sm were evaluated by comparison with in-situ SSH observation data in all fragments of SCAN network. K. Muzalevskiy et al. [100] substantiated the possibility to evaluate the moisture content of bare soil surface with the help of NN, using radar data from Sentinel-1B satellite. Here, NN were considered as a means to construct a regression relationship between the back-scatter coefficients, determined in Sentinel measurements, and polarimetric scatter characteristics, i.e., the anisotropy and entropy, and the reflectivity of SSH. The moisture and roughness of the soil surface are the main characteristics that affect this coefficient [106]). The required moisture values were determined using the obtained data on the reflecting capacity of soil. Estimates of the moisture content of the surface (0–5 cm) soil layer with the use of the proposed method for an agricultural field out of crop with bare soil in a village in Krasnoyarsk Krai, RF, yielded a root-mean-square error ≥3% and a coefficient of determination ≤0.726 [100].

Estimates of Soil Moisture for Drought Conditions

In studies [78, 146], SM was evaluated based on its relationship with the water deficiency index WDI and precipitation. WDI was calculated as a combination of differences between the temperatures of LS and air, obtained with the use of their dependence on the projective cover (LST~VI), determined using MODIS data, in the form of a trapezoid with vortices corresponding to dry and saturated bare soil, vegetation in the state of water stress, and well moistened. These differences were constructed for each pixel using the energy balance equation, where the fluxes of latent and sensible heat fluxes were calculated with the use of data on meteorological characteristics, including air temperature, solar radiation, wind speed, and specific moisture content, the results of calculation of the surface and aerodynamic resistances, as well as MODIS data on albedo and LAI. The calculated values of temperature differences were used to determine the boundaries of a trapezoid for the dry and wet state of LS. The values of meteorological characteristics in [146] were derived from the data of direct measurements at weather stations, and those in [78] were loaded from CLDAS database. At the approach used in the study, the root-mean-square error of soil moisture estimate relative to measurements at different depths varied within 0.067–0.079 m3/m3. Variations in SSH at a global scale over the recent 40 years were analyzed by data ESA CCI (for the layer 0–5 cm), GLDAS-Noah (for the layer 0–10 cm), and ERA-Interim (for the layer 0–7 cm), the latter were closest to the measured values [38]. The authors note an appreciable decrease in the global averages of SSH in 1979–2017 and an increase in the downward trend in 2001–2017, illustrating them by global SSH maps. Drying was observed in sols with the seven main land use types, the largest proportion among them (~80%) being recorded in urban areas. At the same time, in 65% of cases of soil drying, it was due to an increase in temperature, while in 82% of moistening cases, due to the joint effect of temperature, precipitation, and an increase in vegetation area, characterized by NDVI. Under the condition of global warming, the area of soil drying in 1979–2017 increased at a rate of 1% per year. N. Nicolai-Shaw et al. [104] obtained quantitative estimates of the relationship between soil moisture SM during draught, determined by satellite data, and the temperature, precipitation, evapotranspiration, and plant characteristics at a peak of vegetation period for a global scale. The values of SM were obtained by data of ESA-CCI and ERA-Interim/Land; the values of evapotranspiration ET, by GLEAM data, and the values of temperature and precipitation, by data of ERA-Interim (ERA-Tx and ERA-P) (all values with a resolution of 0.25°, at a daily basis, over more than 30 years), the values of NDVI, by AVHRR/NOAA data with a resolution of (1/12)° once in two weeks, the values of LAI and PAR, by MODIS data with a resolution of 1 km once in 8 days, and the data on the degree of coverage by plants, by data of ESA CCI. The values of SM at drought, determined by satellite data from ESA CCI, were compared with analogous values obtained using ERA-Interim data and with the use of LSM and data of multi-frequency radiometer on the precipitation index SPI for the territories of all continents. For different parts of these territories, for drought peaks, specific features of the relationships between SM and the temperature, precipitation, the values of ET and the vegetation area were described. This approach enables monitoring droughts at large space and time scales with the use of satellite data. Note that, to analyze the long-term dynamics of water, energy, and carbon cycles over land surface, the European Space Agency ESA, under its program CCI in 2012, made available the first long-term set of global satellite data on soil moisture SM, ESA CCI SM. This set combined various active and passive microwave products of soil moisture, formed with the use of a single sensor, into three harmonized products—joined active, passive, and combined active + passive microwave products [40]. Compared with the first issue of this set, the latest (as of 2019) version ESA CCI SM contains many improvements along with data of several new satellite sensors with the expansion of its time coverage up to the interval 1978–2018 [53]. W. Wie et al. [148] used ERS data to develop a drought index TVPDI, which, in the authors’ opinion, is a better indicator of land surface drying than those associated with the amount of precipitation P—the temperature LST, and with the state of vegetation—soil moisture SM (because of the interaction between the likely causes of draughts mentioned above). The possibility to use this index was substantiated based on its comparison with data on SM, P, NDVI, LST, other aridity indices, as well as the crop yield per unit area and the net primary production NPP (appropriate analysis was carried out for various space scales). TVPDI estimates at drought were more accurate than the results of the use of P, NDVI, and LST, in this case, the coefficient of correlation between TVPDI and SM was >0.64. TVPDI time series showed a good agreement over the space with other dryness indices and over time, with the crop yield per unit area and NPP for most regions of China. Estimates of TVPDI by MODIS data corresponded also to analogous estimates obtained with the use of Landsat data. As the result, the use of TVPDI for monitoring arid and humid areas in China made it possible to record noticeable space and time differences between the states of dryness–wetness on both monthly and annual scale.

Estimates of Soil Moisture Using New Promising Microwave Satellite Sensors

In the several recent years, integrated (airborne and surface) experiments were carried out to test microwave instruments planned for installation on new space platforms. Thus, study [164] is focused on SSH estimation based on measurement data from an airborne L-range microwave radiometer with synthesized aperture, which carried out earth surface sounding at variable incidence angles and planned for operation on satellites WCOM [135, 136] and TWRS [166]. T. Zhao et al. [164] studied the dependence of SSH estimates on these angles in algorithms corresponding to different polarizations, used to evaluate the brightness temperature and SSH, and in model LPRM; the effect of vegetation and LS roughness on the accuracy of those calculations was also estimated. These results were compared with the data of ground-based measurements of SM and the temperature; in this case, a range of angles was found for different polarizations, making it possible to obtain optimal SSH estimates. Publications [165, 166] present the results of analogous studies carried out to test the instruments for microwave L-range for TWRS satellite and aimed to increase the spatial resolution and the accuracy of SSH mapping by the joint use of the data of active and passive ERS in L-range [166], in L-, C-, and X-ranges [165], as well as data of the optical range. In both studies, SSH estimates for the 0–5 cm layer, obtained with the use of aircraft data of different resolutions (airborne radiometer and radar) for different sounding angles and frequencies, were compared with the results of ground-based measurements of soil moisture and temperature on two grids (in the basins of two rivers of Inner Mongolia (China)). High spatial variability was found in the values of both temperature and SSH for different types of land use (arable land, grain and grass vegetation) [166]. T. Zhao et al. [165] showed that difference between the brightness temperatures for the adjacent frequencies L- and C-, as well as C- and X-ranges are determined by the difference between SSH values, while the difference between the values of the brightness temperature at different angles of incidence are mostly due to the different moisture content of vegetation. In the authors’ opinion, an increase in the number of observation channels leads to more reliable estimates of SSH at a resolution of L-band sensor of about 18 km.

Analysis of Possible Errors in Soil Moisture Estimates by Satellite Data

The difficulty in estimating soil moisture SM by high-resolution SAR data is mostly due to the need to provide accurate parameterization of soil roughness, to take into account the effect of the observation angle, and to ensure the continuity of the data. To use the existing sets of SM estimates by microwave data with a low resolution over a long time, for example, from ESA CCI SM web site, requires the use of scale-reducing algorithms to increase the spatial resolution of estimates from these sets. In addition to the data of such microwave measurements, the results of observations in VNIR ranges were used either to increase the scale of soil moisture estimates by microwave data or to directly determine its values. The advantages of VNIR range measurements are their very high resolution—both spatial (10 m for Sentinel-2) and temporal (less than a day for geostationary satellites) [113]. For example, S. Sabaghy et al. [122] discuss the results of SSH estimation by microwave data of SAR and radiometers of SMAP and SMOS satellites at a decrease in the scale, by data of VNIR ranges Landsat-8 OLI and MODIS, at the use of method of rediscretization (changes in the resolution of images in pixels), by data of airborne measurements with the use of PLMR radiometer and PLIS scatterometer under projects SMAPEx-4,-5 [157] and ground-based measurements in-situ and on OzNet network on the agricultural area Yanko in New South Wales (Australia).

Studying the specifics of errors in SSH estimates with the use of ERS data, carried out in [151], was aimed to improve the methods for optimal assimilation of such estimates in hydrological models. The authors considered the differences between time-invariant and daily errors in SSH estimates by ASCAT and SMAP measurements and analyzed the correlation between these daily errors and a quantitative estimate of plant biomass by leaf area index LAI and the amount of precipitation. In tropical regions, the difference between the time-invariant and daily errors is large in the dry season and small in the rain season. For the major portion of land surface, the daily errors show higher correlation with precipitation than with LAI. In areas with a low vegetation cover, including in barren lands, meadows, and open brushes, rain peaks coincide with peaks in the errors in SSH estimates, and the peaks in LAI values in all cases are attained after the passage of SSH error peaks. Thus, the time variations in the errors in soil moisture estimates is more due to precipitation than to variations in LAI.

Fig. 2.
figure 2

Distribution of moisture storage in the 1-m soil layer in a part of Central Russia on August 22, 2017, calculated by the model (a) with the use of surface measurement data, (b) with initial soil moisture profile specified with the use of ASCAT data, and (c) their difference.

Full size image

An original method for estimating the moisture content of the surface soil layer with the use of GNSS navigation system is proposed in [81]. Taking into account the decrease in the amplitude of the beam refracted by soil, the authors constructed a polynomial regression relationship between the ratio of this amplitude to that of the incident beam and the moisture content of the surface (3.5 cm) layer and its temperature. The form of this relationship was determined by its calibration against measurement data of 18 satellites over 62 days out of the total of 241 days of measurements. The coefficient of correlation was 0.947, and the RMSE value was 0.013 cm3/cm3 (1.3%) at a scatter of SSH values between 0.272 and 0.489 cm3/cm3. This approach seems to require improvement, in particular, carrying out more measurements and detail verifying the accuracy of results.

CONCLUSIONS

The review gives the results of the use of ERS data in different spectral ranges to evaluate the evapotranspiration and the soil surface humidity (SSH) and the moisture reserves of its active layer W as elements of the water and heat regimes of territories of different scales—from local to global. Under such approach, the article presents a retrospective of studies of the processes of water and heat exchange between land areas and the atmosphere (by implementing field hydrological–atmospheric experiments, regular collection of data on the LS and the surface layer of the atmosphere, and the development of models of these processes) since the mid-1980s until now. In addition, a review of databases of ground-based, satellite, and model data (global, regional, and local), accumulated in those studies, is given.

Considerable attention is paid in the study to the results of estimation of SSH and W with the incorporation of microwave range data, including those obtained with the use of SAR, as one of the most effectively developing lines of evaluating those values, which characterize the water availability in the examined territories. The results of application of such data under various physiographic conditions are analyzed. Estimates of the SSH, obtained with the use of neural networks, are given. The issues of the application of ERS data to estimating soil moisture content under drought conditions are considered.

The prospects of the development of satellite technologies for estimating LS and meteorological characteristics, used to calculate soil moisture content and other elements of water and heat regimes, are outlined, including the development of new high-resolution multispectral instrumentation, the construction of large databases, including cloud, methods of their treatment and assimilation in models. A brief description is given to a new promising line of obtaining remote sensing data with the use of numerous microsatellites of small size and weight with high-resolution sensors in either VNIR or microwave ranges. These microsatellites show a high (several times a day) recurrence.

The presented review can be used by researchers focused on simulating the processes of water and heat exchange in territories of different natural zones and obtaining estimates of soil water content and evaporation.