The Impact of Different Phases of the El Niño-Southern Oscillation Phenomenon on Goiás State Rainfalls and Temperature Characteristics Across Three Decades


 Rainfall and temperature are the two key parameters of crop development. Studying the characteristics of these parameters under El Niño-Southern Oscillation (ENSO) conditions is important to better understand the impacts of the different phases of this phenomenon (El Niño, Neutral, and La Niña conditions) on agriculture. This study analyzes 32 years (1980–2011) of climatic data from 128 weather stations across Goiás State in Brazil to determine the behavior of temperature and rainfall time series over three periods (1980–1989; 1990–1999 and 2000–2011) under El Niño, Neutral, and La Niña conditions. The analysis revealed no major impacts of ENSO conditions on accumulated rainfall characteristics, a feature particularly marked in the most recent period (2000–2011). ENSO impacting temperature was identified but presented considerable variability across the periods investigated. These impacts were marked in the first two periods as for maximum temperature and increased from the first to the last period as for minimum temperature. These features were noticed in both analyses in the entire Goiás State and most of the investigated mega-regions, except for the East and Northeast mega-regions as for minimum temperature. There were increases in maximum temperature values throughout the rainfed season (October to March) for all ENSO conditions and investigated periods. Minimum temperature also increased across the three investigated periods, and this was marked in the beginning of the rainfed season (October) under El Niño and Neutral conditions.

Goiás State in Brazil to determine the behavior of temperature and rainfall time series 28 over three periods (1980-1989; 1990-1999 and 2000-2011) under El Niño, Neutral, 29 and La Niña conditions. The analysis revealed no major impacts of ENSO conditions 30 on accumulated rainfall characteristics, a feature particularly marked in the most 31 recent period (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011). ENSO impacting temperature was identified but presented 32 considerable variability across the periods investigated. These impacts were marked 33 in the first two periods as for maximum temperature and increased from the first to  annual rainfall in the study region ranges from 1,130 to 2,040 mm, with a mean of has unique observation times in that interval, hereafter referred to as number of season 127 curve. At this point, we ran a second visual check on the curves and ensured again 128 that data were free of implausible characteristics.
129 Table 1 shows the numbers of season curves for each ENSO phenomenon (La Niña, 130 El Niño, and Neutral) and period (# of season curves), also shown in Figure S1 147 First, we applied a joint analysis considering the 121 weather stations for accumulated 148 rainfalls and temperatures. Then, to confirm the findings obtained in the joint analysis, 149 we disaggregated the weather stations based on the mega-regions of the State of Goiás (Center, East, Northeast, North, and South) ( Figure 1A) and applied the same 151 statistical analysis (described below in section 2.5) for each mega-region.  158 We applied the functional data analysis (FDA) to determine the characteristics and 159 quantify the dissimilarities among functional data of rainfalls and temperatures in the 160 three periods investigated (1980-1989, 1990-1999, and 2000-2011) in the three phases 161 of ENSO. Conceptually, functional data is continuously defined (Ramsay and 162 Silverman 2002) despite being collected in a discrete way, that is, the term 163 "functional" refers to the intrinsic structure of data and not to their form as manifested 164 in observation. In formal terms, the functional record for each individual is defined by 165 ∈ ℕ * pairs ( , ), where is an observation of a realization in and =

166
(1, … , ). Therefore, it is possible to establish a functional relationship through the 167 model = ( ) + in which the noise comes from the measurement process 168 that contributes to the non-smoothness of the observed data = ( 1 , … , The most common smoothing methods to represent the function as a linear

181
The functional data analysis (FDA) results for each ENSO phenomenon phase (La

429
Northeast, North, and South) in the study area, as described in section 2.2.1 ( Figure   430 1A).