RiceAtlas - Questions and proposals for processing

Introduction

This document briefly illustrate the results of exploratory analysis conducted using the PhnoRice algorithm to map rice area and estimate some agro-practices indicator (number of rie seasons, Sowing and harvesting dates), starting from time series of MODIS 250m images

Methods

The PhenoRice algorithm was applied to the

  1. What is “Maturity” ?

Maturity is the length of season. Not clear however what MATURITYS and MATURITYE are. Bhogendra will check.

  1. What is “seed_age” ?

Understood. Information is probably not relevant for our purposes

  1. Where are the areas reported ?

Areas are on the right-most columns of RiceProduction.

Processing needed

To facilitate processing and comparison with PhenoRice results, it would be good if the dataset could be reshaped as follows:

  1. Remove all areas outside of the current PhenoRice analysis

  2. Reshape the DB in a more “processing-friendly” format. I am thinking of something like this:

  • “Location” columns to be kept as is: OBJECTID, ISO, COUNTRY, REGION, SUB_REGION
  • “Location” columns to be added (if/where possible): ISO_SUB, corresponding to the ISO code of the SUB_REGION

  • Reshape the “data” columns to long format. What I’m thinking of is something on these lines:

    OBJECTID ISO … nseas season seas_cat pheno_var start peak end ricearea totarea rice_fc riceprod
    1 VNM … 2 Winter-spring first sow 10 20 30 2500 5000 0.5 2.1
    1 VNM … 2 Summer second sow 100 110 120 1000 5000 0.5 1.1
    1 VNM … 2 NA third sow NA NA NA NA 5000 0.5 NA
    1 VNM … 2 Winter-spring first harv 100 110 120 2500 5000 0.5 2.1
    1 VNM … 2 Summer second harv 190 200 210 1000 5000 0.5 1.1
    1 VNM … 2 NA third harv NA NA NA NA 5000 0.5 NA
    1 VNM … 2 Winter-spring first mat 60 70 80 2500 5000 0.5 2.1
    1 VNM … 2 Summer second mat 150 160 170 1000 5000 0.5 1.1
    1 VNM … 2 NA third mat NA NA NA NA 5000 0.5 NA
    2 VNM … 3 Winter-spring first sow 15 25 35 2500 5000 0.5 3
    2 VNM … 3 Summer second sow 105 115 125 1000 5000 0.5 2
    2 VNM … 3 Off third sow 200 205 210 500 5000 0.5 1

  • This could be done either manually or in R by importing the dataset to a sf object and then using either functionalities from reshape2::melt or tidyr::gather. I’d suggest the second solution, which is less error-prone and recyclable/reproducible (and can be modified if we decide that we need something more/less/different. The function could be added to PhenoriceR package.

  • For Bhogendra: Is it possible for you to work onthis re-sahping ?