Satellite Based Crop Monitoring & Estimation System for Food Security Application in Bangladesh Expert Meeting on Crop Monitoring for Improved Food Security, 17 February 2014, Vientiane, Lao PDR By: Bangladesh Space Research & Remote Sensing Organization (SPARRSO) Scientific Context Food security is an issue of paramount importance. Global climate change with intensification of disasters, Shrinking of agricultural land due to rapidly growing population impose Great challenges - Reliable & up to date crop information is required for addressing the problem
Monthly Rainfall (cm) National Focal Point of Space & RS activities. Research & application of space-based technology to Monitor natural disasters: Flood, Cyclone, River Erosion, Drought etc. Monitor natural resources: Agriculture, Forestry, Fishery, Soil, Water etc. Investigate environmental issues Survey & mapping Provides useful multidisciplinary information to the government. Satellite Ground Stations at SPARRSO 1) NOAA AVHRR, 2) TERRA/AQUA MODIS, 3) MTSAT 4) WINDS Ground Station 5) FY-2D, FY-2E DVB-S System S P A R R S O Calendar of Major Crops in Bangladesh Drought Cyclone Flood JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Aman Rice Boro Rice Wheat JAN FEB MAR APR MAY JUN Potato JUL AUG SEP OCT NOV DEC Sowing Harvest Rainfall
RS-GIS Decision Support System (under development at SPARRSO) RS-GIS decision support system equipped with: 1) Comprehensive geospatial data (satellite-borne and others). 2) RS methodological framework with appropriate algorithm for monitoring of agricultural crops. 3) Protocol for time series geospatial data analysis. 4) Database of spectral signature of surface features. 5) Measurements of temporally evolving biophysical crop parameters. Basic Consideration in Applying RS Technology Consideration A RS data correspond to measured radiation intensities expressed in DN or reflectance values & data do not directly represent any physical surface parameter. Information to be retrieved through numerical approach or by applying classification procedure. Need for effective methodology with optimized data & resource requirement in crop monitoring aspect is well-recognized. Cost effectiveness & sustainability of the methodology is important. Consideration B Size and spatial extent of the crop fields, Spatial heterogeneity of surface features Temporal dynamics of surface features, Spectral contrast between targeted class and other co-existing surface features. Cloud obstacle
Quantitative Information Retrieval by a Two-Step Process 1) Establishing relationship between RS measurements & surface parameters through numeric model or function. ρ ρ i x,t,λ,θ,f(a,b,c,...) - spectral dependency - geometry of observation-illumination t - time x - spatial coordinates 2) Mathematical inversion of the function against satellite data is done to retrieve parameter values. Variability either in spatial, temporal, directional or spectral domain is used to infer information. Schematic Diagram of Crop Monitoring in Bangladesh Separation of vegetated & non-vegetated area Geometric Correction & Georeferencing Masking of forest & homestead vegetation Seasonal vegetation raster layer Spectral characterization of reflectance data Forest & homestead vegetation Atmospheric Correction Monitoring Growth & Condition GCP Collection Biophysical & radiative measurements Spectral Signature Database Ground operation Identification of seasonal rice crop areas Calculation of pixel wise fractional coverage Derivation of total crop area & statistics Temporal characterization of crop radiative response Preparation of raster foundation layer Input: High resolution satellite data Rasterized base maps. m i 1 C A A n j 1 i j Initiative for possible automation. Product Generation
Spectral Response Feature Identification: Spectral Analysis Landsat TM Bands 5,4,3 (RGB), Jan. 21 2013 Water Soil Seasonal Crop 120 120 100 100 80 80 60 60 40 40 20 20 Spectral Response Profile Seasonal Crop Water Soil Home Veg. 0 0 1 2 3 4 5 6 7 Spectral Band Honestead Vegetation Feature Identification: Spatial Analysis Moist sand Ganges R. Dry sand Moist sand area with crop Permanent Vegetation Low land, Bil, Crop Permanent Vegetation Moist Sandy Loam Low land & crop Barind Tract Mainly Aman Crop
Yield Area Strategy Towards Estimation of Rice Crop a) Rice Area Estimation using RS Technology by: SPARRSO Spectral Characterization & Identification Digital Separation RS-GIS Estimation Rice Production b) Rice Yield Estimation using Conventional Field Technique jointly by: Bangladesh Bureau of Statistics (BBS) Having field level network Department of Agriculture Extension (DAE) throughout the country Satellite sensor High Resolution Multispectral Medium Resolution Multispectral Microwave SAR Observational frequency/interval 1 coverage in every 3 years Spatial resolution Utilities 1-5 m GIS spatial vegetation mask for screening out the non-crop vegetation of relatively longer lifetime (other than seasonal) e.g., trees. 3-5 days 25-250 m Analysis of time series satellite data over crop lifetime provide temporal dynamics of radiative responses of seasonal crop cover. Minimum 3 images per season Amplitude, phase, duration & pattern of evolving radiative responses are indicative of crop type. 20-30 m Satellite observation of the Earth surface under cloudy sky condition.
13 13 Spectral Profiling & Biophysical Investigation Selection of profile points at features of interest over satellite image. Extraction of spectral profiles from multidate images. GPS-based positioning & field data collection at selected profile points. TERRA MODIS Image 11 9 12 5 1 10 6 2 3 Road 8 7 4 4 Spectral Profiling Operation
Reflectance Chlorophyll absorption Crop Reflectance 0 20 40 60 80 Crop Height (cm) 10 30 50 70 90 110 Temporal Dynamics for Crop Type Discrimination 0 20 30 40 50 60 70 Days after planting 0 20 30 40 50 60 70 Days after planting December 8 Crop Life Time producing unique temporal pattern of crop reflectance over its life cycle as defined by amplitude, time phase, duration provides indication of crop type & condition Logical Interband Spectral Relational Modeling Computer-based information retrieval considers: Amplitude Phase Crop Calendar Pattern Inter spectral band relationship Spectral characteristics Vegetation Water absorption Wavelength ( m) IF B4 > B3 & B5 > B7 FEATURE TYPE = VEGETATION IF B3 > B4 & B4 < B5 FEATURE TYPE = WATER
Operational Unit 1: Data Preprocssing Operational Unit 2: Analysis & Numerical Operation Operational Unit 3: Ground-based Biophysical Characterization Functional Block Diagram of CEAMONS Operational Unit 1: Data Preprocssing
Crop area (lakh ha) Satellite Derived Aman & Boro Rice Areas in Bangladesh Boro 2012 Aman 2011 Satellite Based Observation of Landcover March 28, 1975 Bare land March 28, 2012 Seasonal crops Bogra Sylhet
Irrigation over large area Crop over previously irrigated areas Feb. 26, 2012 Irrigated area Mar. 28, 2012 Crop area SPARRSO, BBS & DAE the three government organizations of Bangladesh working together for effective crop monitoring in Bangladesh through combined application of RS-GIS and conventional field-based technology. Satellite data of appropriate time frame, coverage & technical specification appear to be a major concern in effective utilization of RS technology for crop monitoring. SPARRSO has been working on an optimizing RS data utilization protocol for operational rice crop monitoring in Bangladesh. A numerical approach has been adopted to address mixed pixel problems in satellite image analysis and it seems to be effective.