Operational Annual wheat area mapping for Afghanistan using Sentinel data and Google Earth Engine

Transcription

1 Operational Annual wheat area mapping for Afghanistan using Sentinel data and Google Earth Engine Presented By: Varun Tiwari Mir Matin International Center for Integrated Mountain Development (ICIMOD)

2 Outline of the Presentation 1. Team 2. Introduction and Objectives 3. Methodology 4. Initial classification results 5. Operationalisation 6. Status 7. Way forward

3 Team Members Mir Matin Faisal Mueen Qamar Haqiq Rahmani Krishna Vadrevu Lee Ellenburg NASA ICIMOD Varun Tiwari Nabin Kumar Yadav Begum Rabeya Rushi MAIL, GIRoA Noorullah Stanikzai & Team

4 Introduction Afghanistan is a land locked country with population of 35 million among which 30% are food insecure Wheat is a major crop and staple food with 80% of total cereal planted area Country is not food sufficient, depended on import

5 Introduction From 2008, MAIL, FEWSNet, WFP conducting pre-harvest survey for area estimation Current estimation is qualitative. More accurate and timely estimation is required for better planning

6 Wheat Area Estimation is Important? Timely and accurate estimation of wheat cropped area is essential for food security management Crop yield forecasting Advance planning Import/Export Pricing control Formulation and implementations of policy related to food procurement Transportation and storage

7 Objective Main objective To develop an operational system for quantitative assessment of wheat sown area to support food security management Specific objective Develop methodology for wheat area mapping at high spatial resolution using Sentinel data Automation of the process and installation at Ministry of Agriculture, Irrigation and Livestock (MAIL), GIRoA Capacity building of MAIL for operation and maintenance of the system

8 Challenges Availability of high resolution satellite datasets. Cloud coverage around Afghanistan. Smaller field size. Limitation of the internet speed for downloading in Afghanistan. Requirement of high end work stations/ Processors/ Computers. Atmospheric correction of datasets.

9 Methodology Crop calendar Foreign Agriculture service office of global analysis IFA Division**

10 Methodology - Overview ```

11 Operationalisation use of Google Earth Engine Advantage Increased speed Data availability Transparency and security Automation

12 Methodology and Results

13 Methodology Data used Sentinel 2A Sentinel 1 Microwave (SAR) Data set used in this study: S1- VV Polarization (for Afghanistan on VV polarization is available) GRD Product.

14 Atmospheric correction of Sentinel 2A data Remote sensing images are contaminated by various radiative process. Absorption by gases Scattering by Aerosols Water Vapours, etc. Approaches for Atmospheric correction Relative Absolute Atmospheric correction done using sen2cor

15 Before Correction Atmospheric Correction After Correction Water Water Urban Urban Vegetation Atmospheric correction using sen2cor Vegetation

16 Methodology Phase 1 S-2 Dataset Agriculture Mask Subset NDVI From Time series NDVI curves for wheat Threshold (for Excluding Other Vegetation features) Mask (Wheat +Fallow land + Other Crops) Result

17 Phase 1 Conti Before Threshold After Threshold

18 Methodology Conti Phase 2 S-2 Dataset Subset Agriculture Mask NDSI>0 (for Fallow) Threshold (for Detecting fallow land) Mask (Fallow land) Result

19 Excluding Fallow from Wheat + Fallow Initial Results It may include some crops

20 Methodology Conti.. Phase 3 Time Series SAR Data Apply Orbital File Thermal Noise Removal Radiometric Calibration In GEE Terrain Correction Filtering GPS for Training GPS for Validation Classification (CART) Wheat Area Validation

21 Time Series SAR Backscatter Laghman Wheat Backscatter Dec Jan Feb March wheat wheat wheat wheat wheat Month wheat wheat wheat wheat wheat

22 Time Series SAR Backscatter 0 Laghman Vegetables Backscatter vegetables vegetables vegetables vegetables -16 Dec Jan Feb March vegetables vegetables vegetables vegetables Month

23 Time Series SAR Backscatter 0 Laghman- Time Series SAR BackScatter Wheat Vegetables Nov Dec Jan Feb March Wheat Vegetables Month

24 Wheat Wheat Vegetation Vegetation

25 Results: Classification Wheat Vegetation

26 Classification

27 Mid Season Accuracy Assessment Row Labels Vegetation Wheat Grand Total User Error Vegetation % Wheat % Grand Total Producer Accuracy 63.16% 78.26% Overall Accuracy

28 Operationalization Downloading Sentinel 2 Images Atmospheric correction Uploading Images To GEE Assert Apply Atmospheric correction Algorithm in GEE Sentinel 1 A Time Series Validation

29 Status and next step Status Initial results verified for four provinces GPS field data collection in progress Way forward More datasets will be used (from April, May, June) to improve the accuracy. Automation of the process and application development Capacity building of MAIL

30 Thank you

31 Time Requirement (Download & Processing) Google Earth Engine Open Source Google Platform Data can be fetched in few Processing (No Cost) can be done Data seconds can be processed and results can be exported in Dual core as RASTER and Vector (even in tablets) formats in few Seconds Traditional Approach Size Proprietary 6 GB = (ESRI, 6X1024= ArcGIS 6144 etc.) MB 6144/2/60 (It has cost) Min High It requires = 51.2Min(Approx.) end work number stations of are required days for (I5,I7 processing. processors) Downloading with 2 Mbps

32 Afghanistan is a land locked country with arid and semi-arid climate. 11% of the arable land mostly lies in temperate ecological zones. Wheat is a major crop and staple food covering 80% of the total cereal planted area in Afghanistan. Climatic conditions such as droughts, increased incidences of pests and crop diseases, lack of irrigation, changing farming practices increased use of pesticides and insecticides, land preparation etc. are some of the factors that have further hampered wheat productivity. Despite being a significant producer, Afghanistan still imports wheat from other countries. The timely forecast/estimation of wheat production is highly important for planning and ensuring food security in case any shortage is predicted.