1 of 26

Satellite data interpretation �over Ghana

Feng Yin

feng.yin.15@ucl.ac.uk

1

2 of 26

EO for agriculture

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

Remote Sensing and Crop Recogniton (wur.nl)

  • Satellite technologies improve the agricultures managements
  • It informs higher-level decision makings through the governments
  • It also informs the decision making for farmers to achieve higher yields and possibly more sustainable practices

​

​

2

3 of 26

EO for agriculture

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

Mapping of peanut crops in Queensland, Australia, using time-series PROBA-V 100-m normalized difference vegetation index imagery (spiedigitallibrary.org)

3

4 of 26

Crop traits: infer from EO (Sentinel-2)

Plant traits

FUNCTIONING

Photosynthesis, C, H2O, yield

​

​

REFLECTANCE

Spectral,

angular

​

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

LAI

Leaf area index

Cab

Chlorophyll

Cm

Dry matter

Cw

Water content

Cb

Leaf senescence

N

Structure

ALA

Leaf angle

Crop status and yield prediction mapping huge importance for EO4 sustainability: EAVs

EO

Plant traits

4

5 of 26

Generate 'typical' (archetype) trait dynamics: #1

​

​

​

​

  • Crop type classification 
      • 🇺🇸 NASS (USDA)
      •       CROME (DEFRA)
      • NE 🇨🇳 You et al., 2021, Nature Sci. Data
      • NCP 🇨🇳 Dong et al., 2020, ESSD
    • Errors and clustering: 60-70% reliable?
  • Derive R for Sentinel 2
    • 2019
    • Atmospheric correction (NCEO ARD)
    • Sample over crop type, area (soil, planting)
  • Estimate P from R 
    • Time series of P with many samples
    • 100,000 per crop, per S2 tile
    • S2 tile 100x100 km2

P|R

US

North China plain

UK

Northeast China

Method

Take Averages over BIG DATA for crops

    • Able to predict clear trait dynamics

Hypothesise 'typical' (archetype) trait

5

6 of 26

Model application:  MC solver

  • Time de-normalisation
  • Hypothesis:
    • Scaling of trait relative to baseline
    • Randomized soil reflectance over time (bounds from all soil samples)
    • Constant soil

​

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Results

Prior distributions:

Ensembles:

At last, model of expectation of optical EO signal

6

7 of 26

Model application: MC solver

Results

  • Compared to ground measurements
  • Producing accurate estimation of time series of bio-physical parameters with limited S2 observations

Field data source: Fang, H., Zhang, Y., Wei, S., Li, W., Ye, Y., Sun, T., & Liu, W. (2019). Validation of global moderate resolution leaf area index (LAI) products over croplands in northeastern China. Remote Sensing of Environment, 233, 111377.

7

8 of 26

Model application: MC solver

  • Compared to ground measurements
  • Producing accurate estimation of time series of bio-physical parameters with limited S2 observations

Results

Field data source: Fang, H., Zhang, Y., Wei, S., Li, W., Ye, Y., Sun, T., & Liu, W. (2019). Validation of global moderate resolution leaf area index (LAI) products over croplands in northeastern China. Remote Sensing of Environment, 233, 111377.

8

9 of 26

Ghana

Author | Email | NCEO Institute

  • High cloud score for most of the time
  • Using only S2 is not enough for early growing season
  • Coarse spatial resolution is not able to pick variation in the fields

MODIS

S2

Planet

9

10 of 26

Ghana: data

Author | Email | NCEO Institute

S2

Planet

10

11 of 26

Ghana: solving with planet

Author | Email | NCEO Institute

https://bit.ly/3JyUUoU

11

12 of 26

Validation: Planet improvement on S2

Author | Email | NCEO Institute

S2 only

Planet only

12

13 of 26

Application: LAI to yield (US)

Author | Email | NCEO Institute

  • Over 800 counties for 2017-2020
  • Weather information only add 1% on top of RS
  • LAI time series explain 85% of r in yield prediction

​

13

14 of 26

Application: LAI to yield (Ghana)

Author | Email | NCEO Institute

14

15 of 26

Summary

Author | Email | NCEO Institute

  • Promising results of using EO for monitoring crop growth status
  • Cloudiness poses the largest issues of using S2 for crop monitoring over Northern Ghana, but Planet seems to be the remedy
  • Encouraging agreements between satellite retrieval with ground measurements
  • Strong linkage between bio-physical parameters and crop final yield
  • Lack of accurate crop classification maps limits the ability to expand the analysis to a larger spatial or a different year

15

16 of 26

Thank you

Author | Email | NCEO Institute

16

17 of 26

EO interpretation: leaf level

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

PROSPECT leaf radiative transfer model and inversion routines • prospect (jbferet.gitlab.io)

MODELING LEAF OPTICAL PROPERTIES (photobiology.info)

17

18 of 26

EO interpretation: Canopy level

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

Philip Lewis. Three-dimensional plant modelling for remote sensing simulation studies using the Botanical Plant Modelling System. Agronomie, EDP Sciences, 1999, 19 (3-4), pp.185-210.

18

19 of 26

EO interpretation : Canopy level

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

https://www.gfz-potsdam.de/sektion/fernerkundung-und-geoinformatik/projekte/enmap/spektrale-modellierung/

https://rami-benchmark.jrc.ec.europa.eu/HTML/RAMI3/MODELS/Drat/Drat.php?blnPrint=true

19

20 of 26

Crop traits: PROSAIL based retrieval

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

R

  • Radiative Transfer model
    • PROSAIL(P): LAI, ALA, Cab, N, Cm, Cw, Cb
    • Soil model (4 parameters S)
    • Simulate Sentinel-2 ,reflectance: R|(P,S)
  • Inverse emulators
    • Estimate (P,S)|R
    • 1-D inverse mapping functions
      • R-> Cab
      • R-> LAI
      • R-> Cm
      • etc

​

20

21 of 26

Ghana: fields explore

Author | Email | NCEO Institute

https://bit.ly/3JyUUoU

21

22 of 26

Spare slides

Author | Email | NCEO Institute

22

23 of 26

Ghana: S2 retrieval

Author | Email | NCEO Institute

  • Lack of S2 observations for the early growing season
  • Used phenology information derived from S1/Planet to provide expectation in generating ensembles

23

24 of 26

Ghana: solving with planet

Author | Email | NCEO Institute

24

25 of 26

Crop traits: Time series

Yin & Lewis | P.lewis@ucl.ac.uk | NCEO-UCL

Need to map traits from EO​

Introduction

Kuester, T., & Spengler, D. (2018). Structural and spectral analysis of cereal canopy reflectance and reflectance anisotropy. Remote Sensing, 10(11), 1767.

25

26 of 26

Author | Email | NCEO Institute

Aims:

    • Show the field measurements and how they look like from satellite
    • Show what we can retrieve from satellite, which are linked to the ground measurements
    • How other sources data can be used to monitor crop in Ghana
    • How these parameters can be used for yield prediction empirically

​

Expectation:

    • Help them to see the possibility of satellite data for crop monitoring
    • Basic concepts of using different sources of satellite data
    • Point out what is lacking to do larger spatial and temporal scale monitoring over Ghana

26