CGIAR-CSI Community Meeting 2020

• Welcome
• Introduction
• Update on community/working group activities
• Shared services
• Workplan review/discussion/approval

As a part of the GARDIAN Ecosystem, the Platform will soon launch the GARDIAN Labs (GLabs), an open collaborative data science platform that allows researchers to work together on the same data science project using datasets securely transferred from GARDIAN and other trusted sources. GLabs will help discoverability, visualization, and analyses of datasets and collaborative analytics using R and Python computer programming languages. GLabs establishes a secure transfer and storage of computer program codes and data files through Globus, another core Shared Service that the Platform provides. Shared programming codes will help accelerate the customization of analytics and avoid possible duplications. This will ultimately contribute to the advances in science by enabling the reproducibility of published work and increased efficiencies. In this session, Sotiris will walk us through GLabs and discuss how best we can use it for our collaborative research work.

The unprecedented combination of high spatial (3-5m) and temporal frequency (1-3 days) offered by Planet data are ideal for observing smallholder farms. Over the last couple of years, the quality of the data improved in many ways but using this data for scientific analysis remains an uphill task due to several reasons, including data access, management, and processing. In this webinar, I will share the experience of processing thousands of Planet tiles to map crop types over large areas. More specifically, I will focus on the following topics: i) understanding Planet Instruments and data catalog, ii) use Planet API to search data, iii) create a pipeline to download and ingest the data as Google Earth Engine asset and iv) process the data in Earth Engine for various agricultural applications including crop type mapping. In addition, I will discuss cloud cover and misregistration issues, and challenges with calibration between different sensors in the Planetscope family.
 
*The work was conducted by the Geospatial Consortium at the UC Davis (https://gfc.ucdavis.edu/) and supported by the Feed the Future Innovation Labs for a) Markets, Risk and Resilience (MRR) and b) Collaborative Research on Sustainable Intensification (SIIL).

Climate Similarity Analysis with R
Julian will introduce the Climate Analogues R package, available at https://github.com/CIAT-DAPA/analogues. He will show how this tool is being used to generate the Extrapolation Domains for climate adaptation options and help us to use the tool in our own analysis.

CSI Representative from each center will give a Center Update presentation (15-minute each) to give a quick overview of the center's geospatial research and an in-depth presentation on a selected study/product.

1. Bioversity-CIAT | Steven Prager
2. CIMMYT | Kai Sonder
3. IITA | Tunrayo Alabi
4. IRRI | Renaud Mathieu

CSI Representative from each center will give a Center Update presentation (15-minute each) to give a quick overview of the center's geospatial research and an in-depth presentation on a selected study/product.

1. IWMI | Kiran Chandrasekharan (Remote)
2. WorldFish | Shwu Jiau Teoh (Remote)
3. ILRI | Francesco Fava & Mike Graham (Remote)
4. ICARDA | Chandrashekhar Biradar (Remote)
5. ICRISAT | Murali Krishna Gumma (Remote)

ECMWF AgERA/ERA5 (Hendrik Boogaard)

ECWMF just published the AgERA5 dataset. It provides daily surface meteorological data for the period 1979-2018 at the spatial resolution of 0.1° grid. AgERA5 ‘connects’ users in the agricultural domain to the ERA5 data set. It includes daily aggregates of agronomic relevant elements, tuned to local day definitions and adapted to the finer topography, finer land use pattern and finer land-sea delineation of the ECMWF HRES operational model. I will explain the background, reasons, methodology and show some results and applications.

WorldClim/BioClim with CMIP6 climate projection (Robert Hijmans)

Ms Sugandh Chauhan, PhD candidate, Natural Resources Department, ITC
Title: Lodging severity assessment using multi-temporal Sentinel-2 data
Crop lodging – the bending of crop stems from the vertical - is a major yield-reducing factor in cereal crops (wheat, maize, rice, barley, oats etc.), with reported losses of up to 80%. The severity of lodging is measured by a lodging score (LS) – an index calculated from the crop angle of inclination (CAI) and crop lodged area percentage. LS is difficult and time consuming to measure manually meaning that information on lodging occurrence and severity is limited and sparse. Remote sensing-based estimates of LS would provide more timely, synoptic and reliable information on crop lodging across vast areas which would improve estimates of crop yield losses, inform insurance loss adjusters and influence management decisions for subsequent seasons. In this context, this study aims to analyze the time-series spectral reflectance of Sentinel-2 data for healthy and lodged wheat plots with different degrees of lodging severity and detect lodging incidence. We conducted the study over 600 ha of wheat fields in the Bonifiche Ferraresi farm, located in Jolanda di Savoia, Ferrara, Italy. We collected 120 samples of crop biophysical parameters (e.g. CAI, crop height) and relevant meteorological data (wind speed and rainfall). The temporal profiles of reflectance, together with continuum removed spectra, were studied for healthy and lodged plots throughout the growth stages. Kruskal Wallis and post hoc Tukey tests were used for further analysis and to test for significant differences among different classes. Our results showed that i) the overall magnitude of reflectance increased with the increase in lodging severity, (ii) red-edge and NIR bands were particularly sensitive in distinguishing healthy from lodged wheat, iii) a blue shift was visible in the red-edge region due to lodging, iv) the change in reflectance caused due to varietal differences or phenological stage was less than that caused due to lodging, and v) with the temporal analysis of Sentinel-2 data, it was possible to detect incidence of the main lodging event. The obtained information is likely to be useful in the context of operational crop lodging severity assessment and food security.

Dr Markus Meyer, Assistant Professor, Natural Resources Department, ITC
Title: Moving beyond farmers – research needs to assess the resilience of food systems
Current changes in food systems in low- and middle-income countries are not only associated with increasing demand for food driven by population growth. Shifts in diets with different production and processing structures likely have a more considerable impact in future: higher land requirements (e.g., more animal-based proteins) or changing production regions and structures (e.g., higher share of (imported) processed food). Whether the expected transition in low- and middle-income countries that is considerably larger than in high income countries is resilient, is questionable.
This contribution aims at highlighting research gaps and at opening potential pathways on how resilience in food systems and associated social and environmental systems could be assessed.
 
Dr Michael Marshall, Assistant Professor, Natural Resources Department, ITC 
Title: “Advancing our understanding of how agroecosystems impact the Earth system”
 

CSI Representative from each center will give a Center Update presentation (15-minute each) to give a quick overview of the center's geospatial research and an in-depth presentation on a selected study/product.

1. ICRAF | Muhammad Ahmed (Remote)
2. IFPRI | Zhe Guo (Remote)
3. CIP | Henry Juarez (Remote)

• "Satellite Data for Smallholder Farmers - the VanderSat case" by Berend de Jong

Good access to resources and opportunities is essential for sustainable development. Improving access, especially in rural areas, requires useful measures of current access to the locations where these resources and opportunities are found.
During the 2020 CGIAR-CSI Community Meeting, Andy Nelson and Rolf de By from ITC/University of Twente, shared some theory on the mapping travel time on a raster surface in R and demonstrated how to generate the input layers and compute travel time to target locations using a precooked example and dataset for Bhutan.
To follow the webinar at your own pace,

1. Install an up-to-date version of R and RStudio
2. Install these packages and their dependencies – raster, rgdal, gdistance – install.packages(“raster”) etc.
3. Download and unzip (or clone) the course material (data, R script and PDF) from https://github.com/gip-itc-nl/accessibility_course_bhutan
4. Download one large SRTM dataset and copy into the /inputs/ directory from point (3) http://ftp.itc.nl/incoming/accessibility/bhutan_dem_srtm.tif

The method is the same as published in the recent article on global travel time mapping and dataset, but a smaller geographic extent was used to make it practical for this session.

Machine learning is the scientific study of algorithms and statistical models that provides the computer the ability to automatically learn and improve from experience. In this workshop, we will work in a set of tools developed by Bioversity-CIAT to facilitate the analysis of metadata and experimental citizen science data, from collating data of different sources, gathering environmental variables, to model selection and visualization. All in a single pipeline in R that can be automated to improve predictions and recommendations for agriculture. The tools are available at https://github.com/agrobioinfoservices (look at the pinned repositories).