The impact of weather variability on wheat and maize production: an improved early warning model for agricultural drought.

Folio: 11190360

Abstract

Global food security is negatively affected by drought. Climate projections show that drought frequency and intensity may increase in different parts of the globe. These increases are particularly hazardous for developing countries. Early season forecasts on drought occurrence and severity could help to better mitigate its negative consequences. Chile is facing what some authors have been calling a mega- drought. Recent studies showed a tendency toward a drier condition (1960-2016) in the central-southern part (30-48°) and indicated that the future projections condition could be potentially underestimated for this region. Recent analyses reveal that 52% of the area under wheat in Chile would face a serious decline in rainfall in 2030-2050. Hence, the country is under urgent need of taking action regarding adaptation measures. The early warning system is the first step for adaptation -if it is possible to anticipate the impact, then it will be feasible to do something to diminish the losses- for which are of increasing need. The understanding of how environmental variables (i.e., precipitation, soil moisture, temperature) impact vegetation productivity under the current climate change will allow improving early warning models. The standard practice for food security assessments at a regional scale is by the monitoring of drought that affects agriculture using vegetation indices (Vis). Such systems offer food security alerts at a medium spatial resolution that are not detailed enough to detect the emergence of local food insecurity conditions and often leverage remote sensing only in a qualitative sense. To account for the impact in yield quantitatively, there is a lack of crop production data (yield, surface, production) worldwide that need to will. Recent improvements in the spatial, spectral and temporal resolutions of Earth Observations satellites (EO) hold significant potential for improved food security assessments. This project aims to develop a near-real-time seasonal prediction model for agricultural drought impact on maize and wheat in central Chile. Were defined three objectives considering wheat and maize: I) select a proxy for biomass from multiple satellite data with similar spectral but different spatial resolution (10m and 250m); 2) develop a biomass estimation model from the proxy derived from public satellite data; and 3) develop a seasonal prediction model for yield from multiple satellite estimates of vegetation, precipitation, soil moisture, and land surface temperature. Throughout the project will be analyzed high spatial resolution satellite data from sensor Sentinel-2 and Landsat 8 OLI by using the new “sen2-Agri” system which allows for ingesting it and retrieving the phenology (start, end, and length of the growing season) and spectral vegetation indices. Also, coarse spatial resolution data from sensor MODIS product MOD13Q1 collection 6 was considered for which will be developed a model to extract phenology and compare it with the one derived from the high-resolution data. For both will be calculated proxies for biomass at different times through the season which will be validated with infield measures of wheat and maize, allowing to develop a biomass model (BM). This BM will be used to obtain yield data by calculating biomass at the end of the season (EOS) and so be able to reconstruct historical records (2000-2022).

The last stage of the project consists in the developing of a prediction model for wheat and maize yield, from satellite environmental data of soil moisture, precipitation, land surface temperature and the proxy of yield itself following a similar approach as in Zambrano et al. (2018). Two models will be tested regardless the availability of soil moisture data, one considering the period 2015-2022 in which all the described variables will be used (M1), and the second by discard soil moisture and defined through for 2000-2022 (M2). The prediction lead-time used will be one to six month before the EOS. It expected that the results of this project could help to anticipate losses on wheat and maize, and so contribute to adaptation measures, and to improves early warning system at a local scale. As the data used here are publicly available globally, the models obtained could be adopted around the globe. Concurso de Proye