Data article - Extreme weather and crop diversification: adaptation to climate change in Brazil
Panel data 1996, 2006, 2017 - Fixed effects Panel data future scenarios Code Stata
Steps to reproduce
This study used as units of observation the Minimum Comparable Areas (MCA) for allowing intertemporal comparisons of the same geographic area, since the number of Brazilian municipalities increased over the years (EHRL, 2017). For this, we made compatible municipalities from the Demographic Censuses from 1980 to 2010, following the methodology proposed by Ehrl (2017). Because MCAs represent the municipality observations, we will simplify the exposition by referring to them as municipalities. The data used for the construction of the Simpson index were extracted from the Agricultural Censuses 1995/1996, 2006, and 2017 from the Brazilian Institute of Geography and Statistics – IBGE. We considered the Gross Value Sold of heads of cattle, pigs, and poultry and the Gross Value Production of horticulture, permanent crops, temporary crops, forestry, and plant extraction at the municipal level. However, several MCAs did not display data on agricultural products at all periods, resulting in an unbalanced panel. From the 1995/1996 Census of Agriculture, data on agricultural products were extracted, which resulted in 3,809 MCAs, while they constituted 3,798 MCAs and 3,813 MCAs in 2006 and 2017, respectively. The daily georeferenced data of maximum and minimum temperature, as well as precipitation, were extracted by the Terrestrial Hydrology Research Group (THRG) (Sheffield; Goteti; Wood, 2006). The database was built by combining global data based on surface observations with the NCEP – NCAR (National Center for Environmental Prediction / National Center for Atmospheric Research) reanalysis. The original data used have a resolution of 0.25º × 0.25º (spatial resolution 28km) of daily precipitation (mm) and daily temperature (C), for the period from 1985 to 2016. However, for the analysis at the MCA level, the data were interpolated to a resolution of 30 meters. It is worth mentioning that the temperature and precipitation data provided by THRG are up to the year 2016. Thus, the impact of the climate did not consider the year 2017. However, the moving averages from five years ago can model the crop choices of the farmers concerning climate variability (Cho; Mccarl, 2017). The data for the socioeconomic, agricultural, and market characteristics of the Brazilian municipalities were also extracted from the Agricultural Censuses 1995/1996, 2006, and 2017. Future climate data were extracted from the General Circulation Models: HadGEM2-ES; MIROC-ESM; and MRI-CGCM3. In the same way as the observed climatic data, the future data used have a resolution of 0.25º × 0.25º of daily precipitation (mm) and daily temperature (C), for the period from 2016 to 2065.