River ice monitoring service () River Ice Monitoring

Main goals:

. To provide information about the ice-covered sections of the river . To visualize the results in graphical form (map), to provide .shp files, .csv files and WMS services . To generate reports in .pdf format . Operational mode - data updated in a few hours after the satellite image acquisition . Fully autonomous service, equipped with modules for automatic downloading and processing of satellite data

End Users:

 Regional Water Management Authorities having competence for surface- and groundwater resources management inside its area of responsibility

 Regional Crisis Management Centres responsible for prediction and prevention of natural disasters and mitigation their consequences

River Ice Monitoring

Operation area The detection of ice is based on Sentinel-1A and Sentinel-1B radar The service will cover the following satellites, that allow to provide user rivers: , , , , , with information with maximum Odra frequency, independently from weather conditions. Data will be downloaded and processed just after acquisition and information about ice/water cover will be uptaded in near real time, accessible via on-line service.

Source data

 Sentinel-1radar images

 IW mode: 3,5 x 22 [m]

 Temporal resolution: 2-5 days

 Polarisation: VV, VH

 Scene size: 250 x 250 [km]

 Sentinel-2 optical images

 ISOK, BDOT10K vector data

Source: PRG (administrative boundaries), BDOO (rivers)

River Ice Monitoring Limitations

Service limitations can be divided into three main groups: 1. Qualitative - the ability to distinguish a limited number of phenomena. View from space with pixel size of Sentinel-1 data allows to divide ice/water cover of the river into four types: water, diffuse ice (frazil ice/floe), Zoom to the single pixels of Sentinel-1 image, visible solid ice cover and ice covered significant difference between by snow lenght and width. 2. Time - timeliness of data depends on further acquisitions. 3. Spatial - the classification efficiency for rivers with a width of 60 meters and more. According to spatial resolution of source image (5x20 meters) between 3 and 12 image pixels can be found within 60 m wide riverbed, depending on stream direction. 04.04.2017, San River (section narrower than 60m) nearby City of Przemyśl The classification results on the river free from ice phenomena, ortophoto: Geoportal 2 River Ice Monitoring Ice phenomena from space

Diffuse ice – floe or frazil ice flowing freely through the riverbed

Solid ice cover – ice cover not covered by snow, also river section with dense cover of fractured/frazil ice

Water – part of the river free from ice phenomena

Snow cover – areas of the riverbed with strong reflectance both in visible and non visible spectrum

Source: Sentinel-2 Source: Sentinel-1 River Ice Monitoring The result of the sample classification

water diffuse ice ice cover snow cover 31.01.2017, Lower Vistula 31.01.2017, Lower Vistula, 31.01.2017, Lower Vistula, Sentinel-2 RGB Sentinel-1 RGB Classification result River Ice Monitoring Service

Algorithm of ice detection and classification is implemented into on-line service, which offers users following functionalities:

 Show classification results as spatial range of ice/water cover on the various (topographic, satellite) map canvas

 Show the most up-to-date Sentinel-1 and Sentinel-2 images as RGB visualisations along the riverbed

 Access to previous classification results by navigation through the past

 Slide and zoom the map by searching specific counties, rivers, places, points of interest

 Use the delivered data in external software/services as exported .shp/.csv layers or WMS link

 Generate statistical report of the ice cover in selected 1-km sections of the riverbed and export it as .pdf file River Ice Monitoring Service Summary

. Ice river monitoring service for selected major rivers will be launched during this winter for RZGW

. The demonstrator has been presented and received well

. After validation in winter it is expected to be technically ready for procurement action, preferably in spring 2018

. Possible further service development: . The possibility of increasing the number of identified ice phenomena types by field interview (winter 2017/2018) . Inclusion of further rivers into the service . The possibility of increasing accuracy by using high resolution TerraSAR-X images

Biomass Products (Poland) Biomass products

Main goals:

1. Seasonal correction of roughness coefficients by analyzing the variability of spectral indicators during the vegetation season in the area of modeling

2. Improved modeling of water flow by monitoring tree felling activities and wild trees/bushes overgrowth between river embankments

End user:

End user of the Biomass Product is the Institute of Meteorology and Water Management – National Research Institute which is responsible for support of: . determination of seasonal changes in roughness coefficient, . improved understanding of the water bodies changes, . improved water flow modelling, . modelling of the flood wave taking into account seasonality. Seasonal biomass changes monitoring

Operation area Product is aimed to investigate connections between Area of interests contain several rivers development of underwater and land surface sections, spread about 500 meters up and vegetation. Confirmation of this relationship would allow to more accurate determination of roughness down along the stream from chosen water coefficient to improve water flow modelling in the gauges: riverbed during all vegetation season. The connection 1. Borkowo on river, between underwater and land vegetation will be 2. Narewka on Narewka river, searched as corelation between seasonal variability of 3. Tchórzew on Tyśmienica river, radiometric indices of the vegetation on the river 4. on Guber river. banks and correction factor determinated for riverbed during the vegetation season.

Source data Biomass changes monitoring is created Wkra Narewka based on the following data sources:

 Sentinel-2 optical images

 Spatial resolution: 10, 20, 60 [m]

 Temporal resolution: 3-5 days  Dependent on weather conditions Tyśmienica Guber  Scene size: 100 x 100 [km]  ISOK, BDOT10K vector data

Seasonal biomass changes monitoring

Tyśmienica - SAVI Mean Selected river sections were analyzed in 4 different buffers: 50, 100, 150 [m] and buffer 1,2 with width twice as the mean river width along 1 area of interest. All cloudless images from vegetation season 2016 and 2017 – around 100 0,8 satellite images were processed. 6 radiometric 0,6 indices: NDVI, IR/R, SAVI, WI, NDWI and STVI were calculated for every river with every 0,4 buffer on every image providing 2352 single rows 0,2 of data. Prepared data have been forwarded to the user 0 for analysis. Initial conclusions indicate, that: 0,0000 0,0500 0,1000 0,1500 0,2000 0,2500

 Observation of growth and atrophy of the Correlation coefficient -0.9105 plants on the river banks could be source for determination of roughness inside Borkowo - NDWI Mean riverbed 1,2  The smaller the buffer, the higher the correlation between state of vegetation and 1 roughness coefficient 0,8  The most precise result were achieved with SAVI index – correlation between spectral 0,6 index and roughness coefficient variability 0,4 was up to 0.9105 0,2  NDWI index could be useful for determination of overgrowing the riverbed 0 with plants (or lack of them) -1,0000 -0,8000 -0,6000 -0,4000 -0,2000 0,0000 Correction factor - 1 Monitoring of felling in floodplains Land cover statistics

Coming years will bring performance of tree Areas selected for piloting the product include selected felling in the valleys of main rivers in Poland. sections of the Vistula river in the middle and lower part. This will significantly change the land cover and will make current data outdated. Therefore, year-by-year land cover classification in floodpains will help to detect both the areas to be cut and areas with extensive vegetation growth. Counting statistics of vegetation along the rivers will help to keep up-to-date input data for modelling flow of potential flood wave.

Land cover statistics Advantages:

. Update of the floodplain ranges by exact data . Performing classification simultaneously for larger areas along longer river section . Generation of land cover statistics broken down into sections consistent with the mileage of the Source: BDOT10K, Source: Sentinel-2 Geoportal 2 river and in various flood wave variants Biomass products Summary

. The EO4EP project will provide by itself the set of Land Use products needed by IMGW for update of hydrological modelling of Poland in 2019

. Experimentation with EO-based information for river flow modelling provided some promising results

. Satellite information about vegetation growth provides good correlation with observed water flow changes (“roughness coefficient”) during the season

. At this stage this solution is not ready for operational use and it will require further R&D activities, but IMGW is interested to use the results to improve modelling of rivers in Poland that are affected by vegetation growth

. This is an activity which is planned as part of the ongoing WBG project Agriculture land types and crops mapping (Arrarat Valley, Armenia)

Agricultural land and crop types mapping

Main goals: . Foster adoption of EO-based technology in the agriculture domain by demonstration of its utility for end-users in Armenia benefiting from World Bank’s technical assistance projects (State Committee of Water Economics, Agriculture Development Fund, National Statistical Service) . Derive information about crops from multi-temporal satellite imagery (Sentinel-1 & 2, Landsat-8) in support of agriculture administration, irrigation management, annual updates and verification of records in farm & crop registry: . Identification of agriculture land types ~ cultivated land (as of growing season 2016 and 2017) . Identification of sown areas of the main crop types or seasonal crop type groups (as of growing season 2016 and 2017) . Aggregation of statistics about areas of derived crops at different levels of administrative units or census tracts . Deploy on-line application to access mapping results Agricultural land and crop types mapping

The service provides seasonal crop mapping products from high resolution satellite imagery. Products provide spatially explicit locations of agriculture land types, sown areas of seasonal crop type groups or, optionally, of individual crop types. In addition, the results including spatially aggregated statistics are accessible via proprietary on-line platform Panther.

Source data Operation area . Multi-temporal composites generated from time series of Sentinel-2 and Landsat-8 optical imagery

. Monthly composites generated from Sentinel-1 SAR data

Sentinel-1 (SAR): May 2017 Sentinel-2 (OPT): May 2017 Ararat Valley (Armenia) Agricultural land types classification Results

Agricultural land types classification (Level-1) product distinguishes three main agriculture land type classes: . arable land (annual crops) . grassland (meadows, pastures…) . permanent cultures (perennial crops) which can be further divided into orchards and vineyards.

Supervised classification has been trained using in-situ reference data from field campaign (courtesy of USAID).

The product is accompanied with a class probability map indicating reliability of the class label assignment by the automated classifier.

Agricultural land types classification (2017)

Probability map (2017) Crop types classification Results

Crop types classification (Level-2) product provides more detailed stratification of the annual crops. The current version of the product includes three general crop groups taking into account predominantly temporal aspect, i.e. the season during which the crop is grown: . Winter crops . Spring crops . Summer crops

Training samples have been collected by visual interpretation of the up-to-date source satellite imagery.

Further improvements of the classification is envisaged by utility of crop type training samples from 2018 spring ground data collection campaign. Seasonal crop type groups classification (2017)

Panther On-line platform for vizualization and assessment of the results

Both Level-1 and Level-2 products are available via the Panther application which provides means to: . On-line exploration of the land-use and crop types classification results – distributions and proportions per customizable spatial units . Assessment of relative shares and absolute figures of agriculture land types‘ areas or sown areas of particular crop types . Calculation of relative structures (%) of the agriculture land/crop types classes at different levels of administrative units . Generating custom indicators (e.g. relative share of permanent crops on total agricultural land) . Visualization of indicators by choropleth maps or different types of charts . Comparison of the calculated statistics across different administrative units . Comparison of the calculated statistics between different growing seasons Agricultural land and crop types mapping Main benefits:

. Monitoring spatial and temporal patterns of agriculture land types and major crop types in the actual growing season. . Products may be updated on annual or sub-seasonal basis . Continuous spatial coverage of the AOI with spatially homogeneous quality . Means to update national crop registry both for upcoming seasons and retrospectively . On-line access to results using intuitive web interface for authorized users

Limitations:

. Availability of the both products and their quality highly depends on quality and completeness of ground reference data . Size of detectable field plots with contiguous crop culture from high resolution data (>1 ha)

Agriculture land types and crops mapping (Kakheti, Georgia)

Agricultural land and crop types mapping

Main goals: . Foster adoption of EO-based technology in the agriculture domain by demonstration of its utility for end-users in Georgia benefiting from World Bank’s technical assistance projects (Ministry of Agriculture, National Wine Agency, FAO Georgia) . Derive information about crops from multi-temporal satellite imagery (Sentinel-1 & 2, Landsat-8) in support of agriculture administration, irrigation management, annual updates and verification of records in farm & crop registry: . Identification of agriculture land types ~ cultivated land (as of growing season 2016 and 2017) . Identification of sown areas of the main crop types or seasonal crop type groups (as of growing season 2016 and 2017) . Aggregation of statistics about areas of derived crops at different levels of administrative units or census tracts . Deploy on-line application to access mapping results Agricultural land and crop types mapping

The service provides seasonal crop mapping products from high resolution satellite imagery. Products provide spatially explicit locations of agriculture land types, sown areas of seasonal crop type groups or, optionally, of individual crop types. In addition, the results including spatially aggregated statistics are accessible via proprietary on-line platform Panther.

Source data Operation area . Multi-temporal composites generated from time series of Sentinel-2 and Landsat-8 optical imagery

. Monthly composites generated from Sentinel-1 SAR data

Sentinel-1 (SAR): May 2017 Sentinel-2 (OPT): May 2017 Kakheti region - Georgia Agricultural land types classification Results

Agricultural land types classification (Level-1) product distinguishes three main agriculture land type classes: . arable land (annual crops) . grassland (meadows, pastures…) . permanent cultures (perennial crops) which are further divided into orchards and vineyards

Supervised classification has been trained using: in-situ reference data for vineyards (courtesy of National Vine Agency), training samples of other classes have been based on visual interpretation of the source satellite imagery.

The product is accompanied with a class probability map indicating reliability of the class label Agricultural land types classification (2017) assignment by the automated classifier. Probability map (2017) Crop types classification Results

Crop types classification (Level-2) product provides more detailed stratification of the annual crops. The current version of the product includes three general crop groups taking into account predominantly temporal aspect, i.e. the season during which the crop is grown: . Winter crops . Spring crops . Summer crops

Training samples have been collected by visual interpretation of the up-to- date source satellite imagery and consulted with local agriculture experts.

Furthermore, map of winter and spring wheat is currently being produced using training samples provided with courtesy of FAO Seasonal crop type groups classification (2017) Georgia. Panther On-line platform for vizualization and assessment of the results

Both Level-1 and Level-2 products are available via the Panther application which provides means to: • On-line exploration of the land-use and crop types classification results – distributions and proportions per customizable spatial units • Assessment of relative shares and absolute figures of agriculture land types‘ areas or sown areas of particular crop types • Calculation of relative structures (%) of the agriculture land/crop types classes at different levels of administrative units • Generating custom indicators (e.g. relative share of permanent crops on total agricultural land) • Visualization of indicators by choropleth maps or different types of charts • Comparison of the calculated statistics across different administrative units • Comparison of the calculated statistics between different growing seasons Agricultural land and crop types mapping

Main benefits:

. Monitoring spatial and temporal patterns of agriculture land types and major crop types in the actual growing season. . Products may be updated on annual or sub-seasonal basis . Continuous spatial coverage of the AOI with spatially homogeneous quality . Means to update national crop registry both for upcoming seasons and retrospectively . On-line access to results using intuitive web interface for authorized users

Limitations:

. Availability of the both products and their quality highly depends on quality and completeness of ground reference data . Size of detectable field plots with contiguous crop culture from high resolution data (>1 ha)

Agriculture productivity monitoring (Ararat Valley, Armenia Kakheti, Georgia) Agriculture productivity monitoring Service

During the EO4SD project we aim to demonstrate that current technology allows to monitor crop productivity from Space, in an on-line, automated and cost-effective manner, at spatial scales from individual fields to whole countries. Focusing on Armenia and Georgia, we specifically target these challenges:

. Tracking overall crop productivity in a region . Predicting yields of the main crops (wheat) . Understanding the impact of weather and climate . Monitoring of patterns of irrigated area . Automated alarms linked to vegetation change along irrigation channels

Agriculture productivity monitoring Service Tracking overall crop productivity

Fine-resulution maps

Evolution through time

Crop classes (GISAT product) Automated stats Vedi region Outcomes: . Year 2017 „the worst recorded” . Follows after exceptionally good 2015 and 2016 . Satellite measurements available before national statistics Agriculture productivity monitoring Service Predicting yields of the main crops

. Generate wheat yield model calibrated with regional yield statistics . Focus area: Dedoplistskaro, Kakheti, Georgia Agriculture productivity monitoring Service Monitoring of patterns of irrigated area (Armenia)

OK year (2002) - June Bad year (2017) - August

Good year (2016) - June Good year (2016) - August Agriculture productivity monitoring Service Understanding the impact of weather and climate

Apart from satellite imagery, the tool brings in available data from ground sensors, notably meteorological stations. Weather mediates crop water demand and in turn influences the need to irrigate. Time-series will help to understand the relative role of weather and irrigation infrastructure.

Meteorology-linked variables to be plotted against satellite-based time-series of vegetation growth Cloudiness

20032017- anomalies2017 Agriculture productivity monitoring Service Automated alarms linked to vegetation change along irrigation channels

Fine-resulution maps

Alarm set up

Vegetation robustness along the channel

Evolution across time DEM production DEM production

Main goals:

The goal is to provide the customer with elevation data for particular areas of interest, validated and enriched on information about technical and operational limitations, quality assessment and quality to price ratios based on analysis conduct in reference to the best datasets being currently in use (ALS).

Information on terrain relief reflected in the form of regular grid of elevation data is the most important source of important inputs in hydrographic modelling.

Highest possible, yet justified by the subject of analysis and computing power applied for the modelling, geometric and hydrographic correctness is the most important factor determining the value of hydrologic modelling results. DEM production Description

Source data Operation area

Input data: 1. Dataset 1. - VHR satellite stereo- Dataset 1.: Area of pairs and/or tri-stereo imageries southern Warsaw’s bundles acquired by Pleiades segment of Vistula river valley. constellation (process of photogrammetric stereo measurement Dataset 1 conducted automatically on the base of Dataset 2.: Selected automatically correlated imageries of buffer area along stereo-pair or tri-stereo). Polish - Belarussian 2. Dataset 2. – pairs of Sentinel-1 and Polish - data applicable for interferometric Ukrainian borders Dataset 2 analysis (interferometric analysis of limited by hydrological division, pairs of Sentinel-1A/B data). but not further than 20 km from borders Input data provided by End-user: (production 1. No data have been provided by the cancelleced due to End-users. End-user might be expected technical limitations of planned product) to provide reference data.

Ancillary input data: 1. High resolution aerial orthophotos and ALS data will be used for ground control points collection. DEM production Results

Methodology of Classification / Measurement Dataset 1: photogrammetric stereo measurement conducted automatically on the base of automatically correlated imageries of stereo-pair or tri-stereo. Spatial Resolution and Coverage Dataset 1: 2m Accuracy, Constraints Vertical accuracy: Dataset 1. (DSM): relative: <1m, absolute: <2m, Planimetric accuracy : Dataset 1. (DSM): <1m,

Accuracy Assessment Approach For each of datasets two main methods of vertical accuracy assessment will be conducted. Both of them will be based on the same reference data – ALS. Availability Dataset 1: 2013 Delivery / Output Format Point cloud: LAS files Raster DEMs: IMG DEM production Summary Application fields: . Basins, sub-basins delimitation – drainage division, watersheds definition (pre-crisis phase), . Hydrologic network definition, verification (pre-crisis phase), . Flood modelling, mainly delimitation of areas endangered with floods - flood-prone areas (pre-crisis phase), . Flood scenario built on the base of different flows – water levels (pre-crisis phase), . Surface roughness maps for flood modelling (pre-crisis phase), . Rapid flood modelling – flood development scenarios (crisis phase), . Selection of places suitable for pumps allocation for water removing (crisis phase), . Delimitation of areas undrained under specific water surface level (crisis phase), . Water depth maps (integration with delimited flood area) – needed for damage assessment (crisis phase), Geomorphologic/hydrologic indexes based on DEM analysis . Topographic normalisation of satellite imageries.

Limitations: . In the case of elevation data derived from the VHR optical satellite data, the vertical and planimetric accuracy will be always lower than the accuracy and level of detail of data elaborated on the base of ALS, yet still on satisfactory level considering foreseen fields of applications. . Capabilities of conducting measurements under vegetation coverage is extremely limited. . In the case of elevation data elaborated on the base of Sentinel-1 data, the accuracy and details content will be much lower than in the case of ALS based data. . Planimetric accuracy and the absolute vertical accuracy could suffer for the areas of interest placed outside of the Polish territory due to lack of reliable ground control points. LULC mapping LULC mapping

Main goals:

The main goal of the product is provision of current information on land cover types coverage as a reference for dynamic phenomena monitoring (land cover changes, anthropopressure, urban sprawl, hydrological network, floods, including pre- and post-crisis phase).

Information on land cover, including hydrographical information and elements of topography build a resource of spatial information widely applicable in crisis management and flood protection practice as well as for the needs of natural protection, agriculture and forestry. These are fields clearly defined as fields of interest for the EO4EP project.

The map is proposed as a reference data applicable in flood prevention as well as agriculture and forestry services scenarios. LULC mapping Description

Source data Operation area

Input data: . Sentinel -2 MSI satellite orthophotos - AOI 1.: Prosna river image enables for the delimitation and catchment. identification of different land cover

classes (like forests, settlements or AOI 2.: Selected AOI 2 waterbodies), area outside of . Base Map for Poland produced internally Polish territory, by GEOSYSTEMS Polska, delimited by the . OpenStreetMap vector databases. Polish – Belarusian border from the west AOI 1 AOI 3 and by the Input data provided by End-user: hydrological division . No data have been provided by the to the east. End-users. AOI 3.: Border Ancillary input data: segment of Bug . High resolution satellite and aerial river valley orthophotomaps will be used to support interpretation process and visual classification enhancement. . Unstructured information sources. LULC mapping Results

Methodology of Classification / Measurement Object-oriented classification, image interpretation (manual refinement), data integration. Spatial Resolution and Coverage Not applicable.

Accuracy, Constraints Thematic accuracy: >90%, Geometric accuracy: <10 m (depends on input Sentinel-2 orthoimageries) Constraints: Minimum Mapping Unit: 0,25 ha Accuracy Assessment Approach Thematic accuracy assessment will be based on a process of quality check conducted independently of the production by qualified interpreters.

Availability 2017

Delivery / Output Format Vector data LULC mapping Summary Application fields:

. Reference data for future land cover monitoring, . Environmental analysis, . Reference data for inundation (flood) detection, . Reference data for flood incidences consequences analysis (after integration with auxiliary statistical data - identification and quantification of endangered population, infrastructure, natural resources yields and real-estate and if possible estimation of replacement value of destroyed yields and real-estates), . Input data for flood risk maps, allowing for integration with external i.e. statistical information sources.

Limitations:

. Main source of input data are Sentinel-2 MSI imageries, being provided as already orthorectified, and the geometrical correction is not foreseen, . The Sentinel-2 data (of 10 m ground resolution) have limited informational content in respect to originally applied for the UA and CLC lev. 4th maps SPOT-5 (with 2,5 m of ground resolution), . Minimum width of linear features is 10 m and in most of cases these features will be integrated from vector data sources. . Vector input data sources for areas placed outside of the Polish territory are being developed independently of the project, since that the technical partners have no influence on their quality and completeness, and some of deficiencies of these resources could not be noticed.