How Reliable are Farmers’ Perceptions about Climate Change? A Case Study in the Upper East Region of Ghana

— The 2014 IPCC report reiterated the importance of local farmers’ perceptions about climate change. A growing number of scientists sup ports that farmers’ in -depth understanding of climate change hazards and their active participation in mitigation actions are key to improving adaptation. This paper attempts to analyze smallholder rice farmers’ perceptions and knowledge about climate chan ge hazards in the Upper East Region of Ghana mainly by looking at the national climatic data and the results of the questionnaire survey we conducted. The climatic data were further analyzed through the Mann-Kendall trend test to find relations between actual rainfall and temperature changes with farmers’ observations. Our analysis on perceptions shows that more than 60% of the respondents experienced climate hazards in the forms of increasing temperature, decreasing rainfall and changing planting time. This result is also supported by the Mann-Kendall trend test. The change in planting time is attributable to the increasing coefficient of variation of the annual rainfall from 16.5% (1996-2005) to 28.1% (2006-2015). It is also due to substantial rainfall deviations within the Region in May, from 1,000 mm in the decade between 1996 and 2005 to 500 mm in the following decade (2006- 2015). We argue that farmers’ observations are largely reliable particularly in observing changes in rainfall patterns. Their observations can also supplement insufficient local meteorological records to better understand local climate change conditions in Western Africa.


INTRODUCTION
The African continent is known to be particularly vulnerable to climate change (Boko et al., 2007). Most parts of Africa have experienced rising near surface temperatures by 0.5 o C or more in the past 50 to 100 years. It is projected that further changes in precipitation, temperature and CO2 level in the atmosphere will cause serious damage to terrestrial ecosystems and agriculture (Niang et al., 2014).
In Ghana, compelling evidence shows temperature increase and rainfall decrease perceptions/awareness about climate change among scientists (IPCC, 2014). Growing evidence supports that farmers' in-depth understanding about climate change hazards is key to improving adaptation (Niang et al., 2014).
Here the fundamental and debatable question arise as to the extent to which farmers' perceptions about climate change are reliable to understand local climatic conditions. To deal with this concern, we attempt to validate farmers' perceptions about climate change by comparing with the actual meteorological data in the same study area for two decades (1996 to 2015).

II.
RESEARCH METHODOLOGY 2.1 Study area This research was conducted in five districts of the Upper East Region, one of three northern regions of Ghana. The region is located at 10.7082° N, 0.9821° W and covers an area of 8,842 km2. The population density is 103 persons/km2. Agriculture is the dominant economic activity here. It employs 80% of the population. It is also the second poorest region in Ghana. Its climate is characterized by two distinct seasons: the wet season from May to October, and the dry season from November to April. Mean annual rainfall in the region ranges from 950 mm to 1,100 mm (Ghana Statistical Service, 2014).
The Upper East Region is one of the major rice producing regions in Ghana. Most recent statistics shows that it accounted for about 21% of rice output in the nation (MOFA, 2016). Non-irrigated rice farming is practiced in all thirteen districts in the region, while irrigated rice farming is practiced in only three districts near Tono and Vea dams. Average farm size per household is 1.3 hectares with a low average yield of 1.8 t/ha (compared with the national average of 2 t/ha). Hoe, cutlass and in some cases bullock ploughs and sickle are mainly used for cultivation.

Sampling
A preliminary survey was carried out to understand the feasibility and significance of the survey for the local participants. Considering the significance of rice production, we selected five out of 13 districts: Bawku West, Binduri, Garu-Tempane, Pusiga and Bawku East. In these districts, we applied stratified random sampling to select sample communities. Using simple random sampling, we selected three representative farming communities in each district and interviewed ten farmers in each community (Table 1). In total, 150 farmers fully participated in our survey. The selected districts have similar characteristics in terms of the climate, soil type, farming system, culture, language and cultivated crops (Ghana Statistical Service, 2014). Due to its similar characteristics, it is mainly known as "Bawku zone."

Data collection and analysis
In the survey, we first attempted to identify sociodemographic characteristics of the respondents partly to understand whether these characteristics influenced their perceptions. We then asked the respondents to describe their observations about long-term (20 years) temperature, rainfall patterns and changes. They were also asked to note if they had made any changes in planting seasons in response to climate change. As most respondents had limited reading and writing skills in English, the questionnaire was interviewer-administered. Also, to overcome language barrier, we translated the questions into the local language. As a result, we had 100% response rate. Survey responses were coded and analyzed in the statistical package for social sciences (SPSS version 20) worksheet. To ascertain climate change patterns in the selected districts, we gathered annual rainfall and temperature data from 1996 to 2015. These data are available at the meteorological department of the Manga Savannah Research Institute (SARI). We used XLSTAT 2017 software to conduct the analysis. We also applied the coefficient of variation (CV) technique to assess the annual and seasonal rainfall variations.
Several . It is a non-parametric trend test for time series data. It enables us to determine whether there is a trend in the time series data or not. Null hypothesis (Ho) is that there is no trend in the data series.
The alternate hypothesis (Ha) is that there is a trend in the data series. The null hypothesis (Ho) is rejected if the pvalue is lower than the significance level. However, when the p-value is higher than the significance level, then we accept the null hypothesis (Ho). The result of the Mann-Kendall test, which shows the statistical significance of the trend in the data set, is usually complemented with the Sen's slope estimator, which denotes the magnitude of the trend. The mathematical representation of the Mann-Kendall trend test is as follows: The nth time series values denoted as: X1, X2, X3…., Xn are replaced in equation (1) below by their relative ranks, R1, R2, R3…., Rn.
where, Should null hypothesis (Ho) is true, then S is approximately normally distributed with: μ = 0 For data sample (n) larger than 10, the standard test statistic Z is computed as the Mann-Kendall test statistic as follows: The presence of a statistically significant trend is evaluated by using the Z value. Positive values of Z indicate an increasing trend, while negative values show decreasing trends. To test for either an increasing or decreasing monotonic (increasing or decreasing) trend at α level of significance, Ho should be rejected if where is obtained from the standard cumulative distribution tables.
The Sen's estimator is determined using the equation (3): where; β is the robust estimate of the trend magnitude. A positive value of β indicates an "upward trend," and a negative value indicates a "downward trend" (Yadav et al., 2014;Rahmat et al., 2015).
The coefficient of variation (CV) explains the deviation in a data series from its central tendencies. The coefficient of variability is found by expressing the standard deviation of a data set as a percentage of its mean value (Bari et al., 2017). Precipitation is the most important parameter that shapes hydrology, water quality and vegetation. If the CV value is high, it depicts a larger spatial variation in the data set (Gajbhiye et al., 2016).

III.
RESULTS AND DISCUSSION 3.1 Socio-demographic characteristics of respondents Regarding the socio-demographic characteristics of the respondents, we found that 61% belonged to an age bracket of 30-49 years old. Male respondents consisted of 64%. However, this does not mean that males dominate rice farming in our study area. Actually, we observed that women were predominant workers in rice farming. In terms of education levels, 81% had no formal education. This constituted 89% of the total female respondents and 76% of the total male respondents. The literacy rate of the female respondents was below the estimated 2018 national literacy level of 76% (Countrymeters, 2018). Despite this social setback, almost 70% of the respondents had more than 11 years of experience in rice farming ( Table 2). Hence their perceptions about long-term climate variability can be overall compatible to the recorded climate data we use.  (Table 3). Overall, nearly all respondents (98%) believed that climate change and weather-related hazards had reduced rice yields.

Table.3: Gender and farmers' perceptions crosstabulation
During our field observation, we found different gender roles in producing rice in the study area. We attempted to see if gender difference affected local climate change perceptions. We conducted a chi-squared analysis and found a significant gender difference regarding drought events (χ2 = 5.038, df = 5 and p = 0.025) ( Table 4). Whereas 72% of the males perceived increasing drought events, 54% of the females did so (Table 3). Although we cannot identify a definite reason to explain this gender disparity, we observed that the male respondents visited their farms more often than their female counterparts largely because the women had to attend to household chores and childcare.

Meteorological data analysis
We examined the extent to which the above findings correspond with the meteorologically recorded rainfall and temperature data in the five districts. The lowest and highest annual rainfall were 671 mm and 1,562 mm between 1996 and 2015. The p-value of 0.048 showed a decreasing trend in annual rainfall in the area ( Table 5). The Sen's slope value of -14.572 suggested that rainfall decreased at a rate of about 15% annually in the past twenty years.
Similarly, the highest and lowest temperature were 37.2 o C and 20.5 o C between 1996 and 2015. The p-value for the maximum temperature of 0.003 means that there existed a trend in the data set. However, the p-value for the minimum temperature of 0.795 shows no trend in the minimum temperature data set ( Table 6). The Sen's slope indicates that the maximum temperature increased by 0.055. It also indicates that the minimum temperature decreased by 0.011. The decrease in minimum temperature, however, was not significant.
Considering the reliability of farmers' perceptions about increasing temperature and drought events, decreasing rainfall and changing planting time  To understand the validity of farmers' responses to climate change by changing planting time, we calculated and examined two decades coefficient of variation in annual rainfalls in two ten-year periods: 1996-2005 and -2006-2015. The results were 16.5% and 28.1%, respectively. This result means that the average annual rainfall in the area varied substantially between 2006 and 2015 by 28%. Also, a graphical plot of the monthly rainfall data in the study area between 1996-2005 and 2006-2015 revealed a substantial shift in the timing and amount of rainfall. In the first decade, maximum attainable rainfall was 1,000 mm whereas the second decade experienced a maximum rainfall of 500 mm in May (Figures 2 and 3). Generally, however, the growing season in this area starts from May and ends in October (Ghana Statistical Service, 2014).

IV.
CONCLUSION This paper has demonstrated that smallholder rice farmers in the Upper East Region of Ghana have carefully observed multiple climate change events and responded to these challenges. More than 60% of the respondents perceived increasing temperature and drought frequency as well as decreasing rainfalls (84%). Nearly all respondents (98%) experienced crop yield reduction due to climaterelated hazards. In response, 82% of our respondents changed their planting time. Their climate observation and remedial actions substantially corresponded with the meteorologically recorded climate data. From 2006 to 2015, the trend of decreasing rainfall in the study area became severe. Rainfall decreased at the rate of about 15% annually. The detected shift in the monthly rainfall decreased from 1,000 mm in the 1996-2005 period to 500 mm in the 2006-2015 period.
Farmers' knowledge about changing climate conditions is based on their careful daily observation. Local farmers, especially experienced ones, can help scientists and policymakers better understand increasingly localized climate events and hazards. Their knowledge can better inform climate-related decision making as well as food security policies in Ghana and Sub-Saharan Africa at large. In turn, policymakers' better understanding about farmers' capacities and needs will make agricultural policies more climate adaptive in the future.