Presentation on theme: "1 Use of High Resolution Satellite Images for Agricultural Land Use Assessment in Romania Dr. G. STANCALIE, Dr. A. MARICA, Dr. E. SAVIN, S.CATANA, C. FLUERARU."— Presentation transcript:
1 Use of High Resolution Satellite Images for Agricultural Land Use Assessment in Romania Dr. G. STANCALIE, Dr. A. MARICA, Dr. E. SAVIN, S.CATANA, C. FLUERARU National Meteorological Administration Bucharest, Romania Workshop on Climatic Analysis and Mapping for Agriculture, Bologna, Italy, 14-17 June 2005
2 Presentation outlook Introduction The Present Romanian Agrometeorological Monitoring System The Improved Integrated Support System for the Agrometeorological Warning and the Identification of the Areas with Agricultural Risk in Romania Useful Satellite Sensors for Agrometeorological Monitoring Satellite – Derived Information for Agricultural Monitoring The Land Cover/Land Use Mapping General Methods for the Generation of the Land Use Map Spatial Resolution Enhancement by Fusion Procedure Multispectral Classification o Unsupervised vs Supervised Classifications for the land use mapping o Improving Classification Using Multi-temporal Images & Multi-sensor Images Statistic Validation of the Land Use Classes Land use mapping based on TERRA/ASTER data Conclusions
3 Introduction The agrometeorological activity undergoing within National Meteorological Administration (NMA), integrates complex issues concerning the current and future evolution of the vegetation state of the crops and water supply of soils with respect to the meteorological parameters evolution, being a particularly important activity whose final objective is to elaborate/edit the agrometeorological bulletins and disseminate information at the level of the decision making factors in agriculture and private farmers.Important infrastructure modernization achievements have been accomplished by the implementation of the National Integrated Meteorological System (SIMIN project) and the development trend of the Romanian agrometeorological monitoring system. Since orbital sensing technologies have undergone unprecedented development, the use of multispectral satellite data in conjunction with traditional means is able to ensure the improvement of the classical determination methods of the agrometeorological parameters, greatly contributing to improve management of agrometeorological phenomena, like the drought. There is a need for accurate agricultural land use assessment; in this respect land use classifications derived from high resolution satellite data proved to be valuable products.
4 The Present Romanian Agrometeorological Monitoring System Processed data (outputs): AGROMONITORING Module of reception, selection and extraction of meteorological data for interested stations and parameters (Synop Programme) Module of high processing of data on information structures (air temperatures, soil temperatures precipitations, evapotranspiration, soil humidity); Module of data storing and manipulation used as input for agrometeorological models - Agrometeorological indices (thermal and hydric) - Monotoring of the vegetation state of the crops - Dynamic of the soil humidity (deficits/excess) - Potential and real evapotranspiration - Elaboration/editing agrometeorological bulletins Server Agrometeorological database Agrometeorological stations - Primary data: - Agro-meteo observations and measurements. - Automatic data pick-up - Data primary processing Meteorological stations Sinoptic/Automatic - Air temperatures min./max. - Precipitations, - Relative humidity - Sunshine duration - Wind speed Real-time monitoring of meteorological and agrometeorological parameters Server Meteorological database (Oracle) DISEMINATION
5 The Improved Integrated Support System for the Agrometeorological Warning and the Identification of the Areas with Agricultural Risk in Romania In the last period an important modernization of the Romanian meteorological infrastructure, based on high technology, has been achieved by the implementation of the National Integrated Meteorological System (SIMIN project), conceived as an integrated system, receiving and processing the data coming from multiple observation sub-systems (surface, radar, satellite sub-systems, etc.) and a lot of types of equipment. Within this project several systems with major significance have been provided: meteorological DOPPLER Radar Network, automatic Weather Network, lightning detection network, satellite reception stations for Meteosat Second Generation, telecommunication networks, visualization system (Nex-REAP) integrating all available information, etc. The main benefit of this investment is relevant for weather forecasting, allowing the detection, surveillance and anticipation of severe meteorological phenomena, which have also an important impact on the agriculture. SIMIN includes a quasi-real time dissemination components of the meteorological information, so that the local authorities and the other users could be promptly informed about the occurrence of severe meteorological events, being able to take, in time, the necessary steps to avoid disasters and damages.
7 The Information Flowchart of the Integrated Support System for Agrometeorological Warning and Identification of Agricultural Hazard Areas Satellite multi-mission receiving station NOAA-AVHRR SPOT-VEGETATION TERRA-MODIS Agrometeorological Stations (fixed and mobile) Automatic meteorological stations Regional Meteorological Centers Subsystem for the management of the agrometeorological database - Main function: data storage, analyzing and data updating. - The input data for the analysis procedures will allow to determine a certain number of hazard indicators of the territory in various agro-climatic conditions. Server Users - strategic - private Subsystem for the management of the thematic, cartographic database -The GIS info-plans and the required information to a structural analysis of the territory. -The GIS database can be personalized and completed through introducing new layers in view to update meta-data associated to each thematic info-plan. Subsystem for displaying the structural vulnerability - Better knowledge of the agroclimatic vulnerability structure; - Crisis management and long term planning having in view the definition of the vulnerability context and its dynamics at various spatial scales (regional, national); - Elaborated results in map-drawing format.
8 Useful Satellite Sensors for Agrometeorological Monitoring NOAA-AVHRR (optical, 4 or 5 channel, broad band scanner, visible to thermal IR, 2400 km swath, 1 km at nadir); Terra/Aqua-MODIS (optical, 250 m, 1 Km, 1330 km swath, 36 spectral bands: an imaging spectrometer, sees every point on Earth every 1-2 days); SPOT/VEGETATION (Optical, 1km, broad band scanner, visible to near IR, spectral bands, synthesis 10-days products); MEDIUM RESOLUTION Landsat-ETM+ 7 spectral bands: 3 – visible, 2 - near IR, 1 – medium IR, 1 – thermal IR, spatial resolution– 30 m, 15 m (PAN only), temporal resolution – 8 days, s wath width -185 km; SPOT 4 Pan and 4 spectral bands: ground resolution: PAN - 10m, green, red, near IR, medium IR – 20m, scene size 60 x 60 Km2; SPOT 5 Pan and 4 spectral bands: ground resolution: PAN - 2.5 & 5m, green, red,near IR, medium IR – 10m, scene size 60 x 60 Km2; Terra/Aqua-ASTER (optical, 10 m, 14 channels, swath of 60 km at nadir, spatial resolution: VNIR-15 m, SWIR-30 m, TIR- 90 m). IRS LISS MODE: 5 spectral bands: 2- visible (green, red), 2 – near and mean IR, 1 bleu synthetic, spatial resolution – 25 m; PAN MODE: 1 spectral band – visible, spatial resolution - 5 m HIGH RESOLUTION
9 Satellite – Derived Information for Agricultural Monitoring Assimilation of remotely sensed data into numerical prediction models (e.g. SWAT, crop models) PredictionMonitoring and early warning Assessment of impacts for extreme meteo events Land use type Intensity and areal extent Use of satellite data as input for crop model yield estimates. Earth Observations from satellites are highly complementary to those collected by in-situ systems. Satellites are often necessary for the provision of synoptic, wide-area coverage and provision of the frequent information required to put in-situ information into broader spatial monitoring of drought conditions.
10 Satellite – Derived Information for Agrometeorology The geo-referenced information, obtained from optical and radar images could be used in determination of certain parameters required for agrometeorology: land surface temperature; evapotranspiration; land (vegetation) albedo; soil moisture; snow cover; fraction of green vegetation, vegetation state; biophysical crop parameters (LAI, FPAR, etc.) land cover land use
11 The Land Cover/Land Use Mapping Requirements for the achievement of the land cover/land use from high resolution images: The structure of this type of information must be at the same time cartographic and statistic; The structure of this type of information must be at the same time cartographic and statistic; It must be suited to be produced at various scales, so as to supply answers adapted to the different decision making levels; It must be suited to be produced at various scales, so as to supply answers adapted to the different decision making levels; Up-dating of this piece of information must be performed fast and easily. Up-dating of this piece of information must be performed fast and easily. The developed methodology implies the following main stages: Preliminary activities for data organizing and selection; Preliminary activities for data organizing and selection; Computer-assisted photo-interpretation and quality control of the obtained results; Computer-assisted photo-interpretation and quality control of the obtained results; Vectorisation of the obtained maps (optional); Vectorisation of the obtained maps (optional); Database validation at the level of the studied geographic area; Database validation at the level of the studied geographic area; Obtaining the final documents, in cartographic, statistic and tabular form. Obtaining the final documents, in cartographic, statistic and tabular form.
12 Satelite Data Processing and Analysis Optical satellite data (LANDSAT–ETM, IRS–PAN/LISS, SPOT- AN/XS, ASTER) have been used to perform the analysis for land use inventory purposes. Optical satellite data (LANDSAT–ETM, IRS–PAN/LISS, SPOT- AN/XS, ASTER) have been used to perform the analysis for land use inventory purposes. A series of specific processing operations for the images were performed, using the ERDAS Imagine and ENVI softwares: A series of specific processing operations for the images were performed, using the ERDAS Imagine and ENVI softwares: Geometric correction and geo-referencing in different map projection system; Geometric correction and geo-referencing in different map projection system; Image improvement (contrast enhancing, slicking, selective contrast, combinations between spectral bands, re-sampling operation); Image improvement (contrast enhancing, slicking, selective contrast, combinations between spectral bands, re-sampling operation); Classifications and grouping; Classifications and grouping; Statistic analyses (for the characterization of classes, the selection of the instructing samples, conceiving classifications). Statistic analyses (for the characterization of classes, the selection of the instructing samples, conceiving classifications).
13 General Methods for the Generation of the Land Use Map Land use map Satellite data: SPOT, IRS, LANDSAT, ASTER
14 Semi-automatic Generation of the Land Use Map Topographic maps Hydrographic networkAdministrative boundaries Available land use in situ data IRS PAN+ LISS SPOT PAN +XS IRS PAN+LANDSAT ETM SPOT PAN + ASTER HR image product Geometric correction Subset/Mosaic Radiometric enhancement Pre-processed HR image product Interpretation Classifications Map editing Land use map
16 Exemple of fusion for IRS images IRS Pan Image Fine geometrical resolution (5 m) IRS Pan+LISS fusion image (5m) Advantages: Fine geometrical resolution Rich multispectral information IRS LISS image Rich multispectral information (25 m)
17 Statistics of the Land Use Classes Land use classesSurface(km2) winter crop42.0 summer crop57.8 uncultivated soil24.9 forest590.2 bare soil121.7 water10.7 villages152.2 constructions19.6 pastures, vineyards424.2 Total1443.3 IRS derived land use map (Arges basin)
18 Multispectral Classification Pixel based classification Pixel based classification Non-supervised classification – n classes Regrouping – following interpretation rules Emphases cultivated zones Supervised classification based on training areas Regrouping Parcel based classification Parcel based classification Uses neo-channels (derived from PAN) : texture and variance Non-supervised classification based on dynamic clusters (mobile centers) Emphases cultivated areas vs. urban zones Regrouping Combination of the two type of classifications Combination of the two type of classifications Image segmentation Image segmentation in order to identify homogenous areas (parcels) for radiometric and textural point of view.
19 The result of pixels based classification The result of area based classification Result of the combined classifications Winter crops Summer crops Pastures, vineyards Urban zones Urban zones (big building) water uncultivated forest Bare soil Land use/cover map
20 IRS PAN/LISS - Derived Land Use Western Romanian Plain (Crisul Alb basin)
21 Statistic Validation of the Land Use Classes Histograms associated to land use classes in the radiometric channels of the IRS data Classes discrimination test using the separability cell array
22 Land use mapping based on TERRA/ASTER data 1) Geo-referencing of the ASTER data The images have been co-registered to WGS84 datum and UTM projection zone 34 and rotated with different angles in order to co-locate and analyzed with topographic maps. 2) Detection of cloud and water Water, clouds and cloud shadows have been separated and the land patterns have been emphasized for the classification procedures. 3) Data classification procedures Unsupervised classification: find the right number of classes with a specified number of iterations. Supervised classification: based on training areas using a priori knowledge of the number of classes, as well as knowledge concerning statistical aspects of the classes. Regrouping 4) Correction After classifications, corrections have been applied in order to eliminate the isolated pixels. 5)Validation of results Validation included checking methods based on in situsampling focused on the classification precision, the exactness of the geographic boundaries and the homogeneousness of the occupation structure. 6) Integration in a GIS environment allowing obtaining thematic maps at different scales.
23 ASTER satellite image, 15 m resolution, Crisul Alb basin (subset zone) Un-supervised classification, raster format Un-supervised classification converted in vector format Land Use Obtained by Un-supervised Classification of TERRA/ASTER Data
24 Land Use Obtained by Un-supervised Classification of TERRA/ASTER Data (cont.) A – image resulted after Conversion Raster to Vector AB B – image resulted after applying a Majority Filter (Neighboroght Statistics) In order to include the pixels that belong to very small polygons into the land use units a special procedure has been applied using the Neighbourhood Statistics function from Arc Map software. The result was a simplified classification in raster format. This raster was converted in vector format by Convert Raster to Features (GRID) of Arc Map software.
25 Supervised classification in 7 land use classes + urbanclass urban ASTER image – color composite Land Use Obtained from Supervised Classification Based on Training Areas
26 Unsupervised vs Supervised Classifications for the land use mapping Unsupervised classification: 7 classes + urban obtained by merging 50 classes ASTER image
27 Land Use Obtained from TERRA/ASTER Data (Detail)
28 Improving Classification Using Multi-temporal Images In classifying satellite data different land use show similar spectral signatures during the vegetation period. This is especially true for grassland of different use and cereals of different stages of maturity. Due to the presence of different plant communities and varying forms of cultivation grassland can in general not be separated from cereals in the feature space using only one image. The spectral differences within grassland could be higher than the differences between grassland and cereals or other crops. Therefore multi-temporal images should be used to improve separability.
29 Use of Multi-temporal & Multi-sensor Images for Crop Identifications SPOT4 XS 18 April 2000IRS-LISS 4 August 2000
30 Use of Multi-temporal & Multi-sensor Images for Crop Identifications ASTER – September.2003SPOT 4 – April 2000IRS-LISS – August 2000
31 Conclusions The new investments in the Romanian meteorological integrated system will contribute to improving the monitoring system of the agrometeorological parameters, using highly efficient methodologies and techniques (mathematical modeling, GIS, remote sensing), in order to evaluate the vegetation state of the crops, of the moisture and soil water deficit dynamics, for the optimization of the agricultural management. The results, such as the agrometeorological bulletins (diagnosis and forecasts) will be disseminate via Internet at different customers including the decision making factors in agriculture, extension services, insurance companies, farmers, media, etc. The high resolution satellite data proved to be useful for the agricultural land use assessment being tailored to multiscale requirements. Covering wide areas satellite imagery comes in a full range of resolutions from 20 m down to 2.5 m, for work on regional or local scales (from 1:100 000 to 1:10 000). These images are also useful for observing and analysing the evolution of land surfaces to try and understand changes affecting vast areas or precise locations.