Project scheduled for 2018.
Eliminate hunger and eradicate poverty around the world are key priorities of the UN’s Sustainable Development Goals. To accomplish these arduous tasks, efforts should be made to improve the productivity of smallholder farmers, who represent the weakest link in the agricultural production system. High resolution yield maps emerge as crucial to strengthen the analysis and decision making by policy-makers and to optimize the design and implementation of agro-management support policies and aid packages supplied by NGO’s and other relevant stakeholders. However, currently available cropland products suffer from major limitations, such as, absence of precise spatial location of the cropped areas, coarse resolution or limited differentiation of crop types and cropping intensities. This lack of high quality data can be ameliorated with the use of recently developed, inexpensive, satellite-based techniques. Being this problematic traditionally analyzed in African and Asian countries, the present research proposal focuses on the Dry Corridor, a region in Central America whose population remain extremely vulnerable to food security problems as a result of sustained crop exposure to droughts.
The Dry Corridor is an imprecise geographic zone in Central America—stretching from Guatemala to Costa Rica and Panama—with homogeneous agroclimate, ecosystems, and livelihoods. Households in the Dry Corridor are exposed to poverty, climate change, extreme climate events, violence and food insecurity. Around 60\% of the rural population from a total of 10.5 million living in this dry tropic area live in poverty and lack adequate livelihoods. Based on the projected population increase for this area, the basic household, strongly dependent on agricultural self-production, will experience an increase in poverty conditions and their vulnerability to food crises. The action of climate change will presumably favor the intensification of the latter two processes.
According to FAO, 30% of Central America is located within the Dry Corridor, 60% of which suffers from drought conditions classified as ‘elevated’ and ‘severe’. Droughts in the Dry Corridor are cyclical and narrowly connected to El Niño Southern Oscillation (ENSO). Rainy seasons, from May-June to November, feature a bimodal pattern favoring multiple cropping over the season, with a first (primera), second (postrera) and, occasionally, a late (apante) season. Droughts are typically manifested as precipitation anomalies within the rainy season, especially on their onset and around the summer heat wave. Two consecutive years of drought from 2015 to mid-2016, linked to a powerful ENSO, decimated crop harvests. The typical farm has an average size of 0.35 hectares, dedicated to maize or bean. Traditional agro-management practices are employed throughout and the use of fertilizers is scarce.
Case study: Honduras
The Corredor Seco Food Security Project (PROSASUR) is part of the Alianza para el Corredor Seco (Alliance for the Dry Corridor, ACS), which is the flagship program of the Government of Honduras for food and nutritional security in a large strip of land characterized by being under severe dry seasons. TThis project is designed to give attention to 12,000 families in 25 municipalities in the departments of Francisco Morazán, El Paraíso and Choluteca. Detailed geo-referenced yield and agro-management farm data for this area is available from 2016. Timely yield maps of this region will help to distribute food assistance and support packages (seeds and fertilizers) efficiently after the primera season.
I will use linear statistical crop models fed with the PROSASUR yield data, high resolution Vegetation Indices (VIs) and drought-specific indicators. I will construct VIs anomalies from Landsat 8 (30 meters) using Google Earth Engine (GEE). Total evapotranspiration, crucial to determine yield performance in water-limited environments, will be obtained from EEFlux (GEE-based). My methodology will also account for area-specific characteristics, such as the effect of ENSO on yields (described below).
Land cover classification. Using Landsat’s Archive and recent tools to differentiate crops based on satellite imagery (Azzari and Lobell, 2017), a supervised classification algorithm based in Random Forest will be implemented in GEE to distinguish maize and bean farms. The algorithm will also be fed with soil and geophysical characteristics derived from the PREVDA Atlas. Image classification will be validated using different techniques, such as visual identification, field-level data and aggregate statistics from the Honduran Government.
ENSO. I will account for the effect of ENSO on yields. This phenomenon is characterized by an abnormal heating of the Sea Surface Temperature (SST) over the equatorial strip of the Pacific Ocean. The use of an ENSO index from AVHRR remote-sensing SST data (Song, 2016) as an additional driver will increase yield’s explained variance, given its influence on the studied area.
I envisage two natural lines of extension of this project: a) temporal, by which yield forecasts could be anticipated on the basis of the empirical models estimated in the previous section and the use of precipitation seasonal forecasts for this area together with near real time VI data; b) spatial, broadening the study area to the Dry Corridor’s CA-4 area (Guatemala, El Salvador, Honduras and Nicaragua) using a scalable uncalibrated methodology (Jin et al, 2017) —based on crop models, such as APSIM for maize and DSSAT-BEANGRO (Eitzinger et al, 2017) for beans.