Abstract

Farmers in sub-Saharan Africa face many difficulties when making farming decisions due to unexpected changes in weather and climate. Access to hydroclimatic information can potentially assist farmers to adapt. This study explores the extent to which seasonal climate forecasts can meet hydroclimatic information needs of rice farmers in northern Ghana. First, 62 rice farmers across 12 communities were interviewed about their information needs. Results showed that importance of hydroclimatic information depends on the frequency of use and farming type (rain-fed, irrigated, or both). Generally, farmers perceived rainfall distribution, dam water level, and temperature as very important information, followed by total rainfall amount and onset ranked as important. These findings informed our skills assessment of rainfall (Prcp), minimum temperature (Tmin), and maximum temperature (Tmax) from the European Centre for Medium-Range Weather Forecasts (ECMWF-S4) and at lead times of 0 to 2 months. Forecast bias, correlation, and skills for all variables vary with season and location but are generally unsystematic and relatively constant with forecast lead time. Making it possible to meet farmers’ needs at their most preferred lead time of 1 month before the farming season. ECMWF-S4 exhibited skill in Prcp, Tmin, and Tmax in northern Ghana except for a few grid cells in MAM for Prcp and SON for Tmin and Tmax. Tmin and Tmax forecasts were more skillful than Prcp. We conclude that the participatory coproduction approach used in this study provides better insight for understanding demand-driven climate information services and that the ECMWF-S4 seasonal forecast system has the potential to provide actionable hydroclimatic information that may support farmers’ decisions.

1. Introduction

The agriculture sector of many West African countries has yet to realize its full production potential in terms of agricultural yield. Compared to levels achieved in the 1960s, the sector is considered to be underperforming (Nin-Pratt et al. 2011; Benin et al. 2011). The low performance of the sector has many causes, including political and institutional constraints, low adoption rate of sociotechnical innovations, and biophysical factors, including highly variable climatic conditions (Baltzer and Hansen 2011). Climate variability in large parts of Africa is projected to increase due to global warming (Niang et al. 2014; Salack et al. 2016), which is likely to have severe impacts on the agricultural production (Rockström et al. 2014; Schlenker and Lobell 2010). This is particularly the case for sub-Saharan Africa where smallholder farmers largely depend on rain-fed agriculture and small-scale irrigation systems. Changing rainfall patterns could necessitate significant adjustments to farming activities (Salack et al. 2015). For example, changes in the onset, duration, and end of the rainy seasons have already affected planting patterns and the farming calendar (Ambani and Fiona 2014).

In Ghana, significant changes in farm activities caused by climate variability and change are already evident and efforts to manage the negative effects of this change on agricultural production have had limited success. Water scarcity and reliance on unpredictable rainfall remain major factors limiting crop production in the country. One of the most important concerns in this regard is the increasing rice yields (Donkoh et al. 2010; Kranjac-Berisavljevic et al. 2003). Rice is currently a key staple crop in Ghana for which the consumption has increased in recent years (Mabe et al. 2012). As a result, the production of rice needs to increase to meet rising demands under the increasing variable climatic conditions (SARI 2011). This poses a significant challenge, as farmers have to make several climate-sensitive decisions months in advance of the rice farming season (Asante and Amuakwa-Mensah 2015). A similar challenge exists in irrigated rice farming. The difficulty of predicting rainfall and consequently river discharge affects the decisions of water managers about water distribution to the irrigated farmlands. The use of weather and climate forecasts could be an instrument that helps farmers in their decision-making to improve agricultural productivity and food security (Hansen et al. 2009).

Previous research on hydroclimatic information to support farmers in their decision-making can be broadly divided into two directions. The first is social science studies that explored in a mostly bottom-up fashion the weather and climate forecast information needs of smallholder farmers and potential challenges they encounter. Results showed that farmers do receive weather and climate information, mainly through radios and local administration (Feleke 2015). Relatively few farmers find the information useful in their operational decision-making. Language problems, difficulty in understanding forecast terminology, and inconsistency in the time of information provision constrain farmers in the use of weather and climate information (Feleke 2015). Other studies conclude that weather and climate information currently received by farmers is insufficient and service improvements are needed to make better use of the available weather and climate forecasts for informed decision-making (Onyango et al. 2014).

The second line of research focuses on technical and top-down approaches assessing the skills of existing forecasts for several regions across the globe (Ogutu et al. 2017; Barnston et al. 2010; Kumar et al. 2001). These studies often conclude that weather and climate forecasts have considerable potential to improve agricultural management and rural livelihoods (Ouédraogo et al. 2015; Roudier et al. 2014; Hansen et al. 2009) but do not connect it to the needs of farmers to make informed decisions. Several forecasting systems have been developed and used (e.g., Alves et al. 2002; Kanamitsu et al. 2002; Mason et al. 1999; Stockdale et al. 1998). The European Centre for Medium-Range Weather Forecasts System 4 (ECMWF-S4) ensemble seasonal climate forecasting system is a state-of-the-art system with an ensemble of 15 members found to be skillful in many regions across the globe. Several authors have argued that it has potential value for providing climate services for vulnerable sectors including agriculture, energy, and health (GFCS 2016; Manzanas et al. 2012). Nonetheless, until now only a few studies have used ECMWF-S4 for Africa (Ogutu et al. 2017; Trambauer et al. 2015; QWeCI 2013).

In this study, we aim to connect these two different lines of research to gain insights into demand-driven climate service for rice farmers’ adaptive decision-making. More specifically, we explore if and how seasonal climate forecasts of the ECMWF-S4 can meet the hydroclimatic information needs of rice farmers in northern Ghana. To do this, we used social science methods (interviews, workshops) combined with a skills assessment of ECMWF-S4 seasonal climate forecast system. Therefore, to meet the main objective, we implement a two-step approach by 1) identifying the information needs and 2) assessing hindcast skills (verification).

The paper proceeds as follows. First, we briefly introduce the case study region. In section 3, we describe the methods used for data collection and analysis, followed by section 4 where we present the findings of the farmers’ needs assessment and the performance evaluation of the forecast. We discuss the findings in section 5, followed by a concluding section.

2. Study area

The northern area of Ghana is located within the intertropical convergence zone (ITCZ) where the movement of the two air masses, the Harmattan or northeast trade winds and the southwest monsoon winds, determines the nature of the climate (Liebe 2002). The area is associated with erratic unimodal rainfall with total annual precipitation ranging from 400 to 1200 mm. The north of Ghana has a challenging climatic condition such as a long dry season of about 6 to 7 months followed by a 5-month rainy season (April/May to September/October). The area is characterized by frequently occurring drought and flood events (Amikuzuno and Donkoh 2012; Asare-Kyei et al. 2015). Temperatures in this part of Ghana are higher compared to the southern part of the country. Maximum temperatures range from 26°C in August to 40°C in March or April (Mdemu 2008). This makes agriculture activities here highly vulnerable to climate variability and change.

Northern Ghana comprises the upper west region, the upper east region, and the northern region (Runge-Metzger and Diehl 1993). The poverty level of northern Ghana is higher compared to the southern regions; even after over 30 years of agricultural-led development projects, the northern regions of Ghana remain impoverished (IFAD 2012; Morris et al. 1999). According to a recent report from Ghana Statistical Service, about 80% of the economically active population in this part of Ghana engages in agriculture (GSS 2013). The main crops are rice, maize, soybean, millet, cassava guinea-corn, groundnut, beans, and sorghum, with some farmers also producing dry season tomatoes, pepper, cabbage, and onions mainly for consumption with surpluses for the market (GSS 2013). Generally, average farmland size varies with crop type, being about 0.27 hectares (ha) for soybean, 0.72 ha for rice, and 1.06 ha for maize. Rice production in the area declined from 3.20 megatons (MT) ha−1 in 2010 to 2.32 MT ha−1 in 2015 despite an increasing demand (USAID 2017). The period for rice farming thus has information needs that are similar across the three regions of northern Ghana because of similar agro-ecological conditions, even though there are individual preferences for different rice varieties based on production rational (GIDA 2016).

To mitigate irregular water availability for farming and domestic activities in the northern region, about 20 small and large irrigation schemes have been developed with the Bontanga irrigation scheme being the largest in the Kumbungu District (Fig. 1). The Bontanga irrigation scheme sources its water from the Bontanga River, a tributary of the White Volta River. The scheme has a potential area of 800 ha but 450 ha are currently irrigated. Out of this, 240 ha is used for lowland rice cultivation. In 2016, the scheme included about 600 farmers (~100 women and ~500 men) from 13 different communities with an average of 0.8 ha per farmer. They engaged in rain-fed and irrigated rice farming in the rain and dry season respectively (GIDA 2011, 2016).

Fig. 1.

(a) Northern Ghana in a black rectangle relative to (b) Africa showing Ghana. The pink circle shows the position of Bontanga river and irrigation dam

Fig. 1.

(a) Northern Ghana in a black rectangle relative to (b) Africa showing Ghana. The pink circle shows the position of Bontanga river and irrigation dam

3. Research methodology and data

This research was conducted in three main steps. First, document analyses, interviews (n = 62), and a feedback workshop were used to obtain information about the hydroclimatic information needs of rice farmers in the communities around the Bontanga irrigation scheme. In the second step, we evaluated the skills of the European Centre for Medium-Range Weather Forecasts System 4 ensemble seasonal climate forecasts using probabilistic verification statistics. Third, we assessed the potential for meeting the hydroclimatic information needs of farmers at their expected lead time.

Data collection and analysis

1) Assessment of hydroclimatic information needs

To collect data on hydroclimatic information needs for farmers’ decision-making we designed a structured interview guide based on document analysis (Bowen 2009) and previous studies (Roudier et al. 2014; Crane et al. 2010; Roncoli et al. 2009). The interview protocol covered both open and closed questions on (i) respondents’ general perception of climate variability and change, (ii) hydroclimatic information needs for decision-making where farmers could identify their hydroclimatic information requirements in each stage of the farming process, and (iii) general information about respondents (see Table S3 in the online supplemental material for the interview guide). The interview guide was pilot-tested twice to ensure that the questions were understandable and unambiguous.

In total 62 rice farmers were interviewed (Table 1). Each interview lasted for about 30–40 min and was audio recorded. In the sampling process, we aimed to balance between types of farmers—irrigation only (IO), rain-fed only (RO), and both irrigated and rain-fed (BIR)—and their location within the irrigation scheme (up-, mid- and downstream of the Bontanga River). Individual farmers were selected based on their rice farming experience (more than 5 years) and their willingness to participate in the survey. We included IO farmers (n = 11), RO farmers (n = 20), and BIR farmers (n = 31). After completing the interviews and processing the data, a one-day feedback workshop was organized to discuss and validate the interview results with representatives from each of the 12 selected communities. The aim was to reduce interpretation bias by the researchers, to collectively rank information needs, to improve understanding of the respondents needs, to share key insights of the research team, and to identify farmers for follow-up studies. The data and information gathered from the interviews and workshop were analyzed using descriptive statistics (frequency and percentage). The analyzed demography and farming characteristics of the farmers are presented in Table 1 showing frequencies and percentages.

Table 1.

Sociodemographic structure of respondents (n = 62).

Sociodemographic structure of respondents (n = 62).
Sociodemographic structure of respondents (n = 62).

2) Seasonal climate forecast verification

(i) Data collection

Thirty years of daily hindcast data of total precipitation (Prcp), minimum temperature (Tmin), and maximum temperature (Tmax) were collected from ECMWF-S4. The data are from an ensemble of 15 members at approximately 0.75° horizontal resolution. The data initialization used for this analysis starts on the first day of every month from 1981 to 2010. Each of the 15 ensemble members provides forecasts of up to 7 months. Also, 30 years (1981 to 2010) of Water and Global Change (WATCH) forcing data and ERA-Interim daily data (jointly referred to as WFDEI) of the same variables (Prcp, Tmin, and Tmax) were used as reference observation (Weedon et al. 2014) because of the sparse network of weather stations in the area and because the quality of data available does not allow for proper spatial validation.

WFDEI has been considered useful for evaluation purposes in East Africa (Ogutu et al. 2017). An intercomparison analysis of precipitation variability and trends in Ghana has shown that GPCC, which is an input dataset of WFDEI, performed well when compared to Ghana Meteorological Agency (GMET) station data for monthly totals in northern Ghana (Manzanas et al. 2014). We performed further analysis of Prcp, Tmin, and Tmax to recognize the extent of the existing bias on a daily time scale (see Figs. S1 and S2 in the online supplemental material). But the results of this validation were not very encouraging as WFDEI could not properly estimate the variables at a daily time scale even though it could capture well the temporal trend of variability of the variables. Comparing point data from a wide grid to GMET station data could therefore have affected the results.

(ii) Data analysis

This study uses two well-documented verification measures [the generalized discrimination score (GDS) and relative operating curve skill score (ROCSS)] to assess the performance of the forecast to a standard reference (i.e., the climatological forecasts and observed climatology). Indicator values for these measures can be zero, denoting a forecast being as good as the reference, or positive or negative, implying an improvement and no skill respectively.

In more detail, the forecast verification is performed for three different periods of the rainy seasons of northern Ghana: MAM (March–May, coinciding with onset), JJA (June–August, for peak monsoon season), and SON (September–November, for cessation) (Amekudzi et al. 2015; Sultan and Janicot 2003). Our results from step 1 showed that farmers preferred lead time ranges between 0 and 2 months (see section 4a). The skill was verified at 0-, 1-, and 2-month lead times corresponding to the months the forecast started before a growing season. The verification was carried out on the ensemble mean of all members as the accuracy of the verification improves with larger ensemble size. Large ensembles are particularly important if extreme events are to be forecasted (Ferro et al. 2008; Weigel et al. 2007). Prior to the validation, we matched forecast data spatial resolutions (0.75°) to observe data resolution (0.5°) using bilinear interpolation, which is a widely used method in climate forecast validation exercises (Cofiño et al. 2018; Ogutu et al. 2017; Bedia and Iturbide 2017). Three-monthly averages of the forecasts and observations were computed to allow the validation scores on seasonal time scale. The evaluation was carried out at gridpoint level and for three regions within northern Ghana where rainfall patterns are similar (Nkrumah et al. 2014). We analyzed mean biases for each of the three seasons and for different lead times (Willmott et al. 2012). The strength of the relationship between the ensemble mean and the verifying observations was assessed using Spearman’s rank correlation coefficient.

The generalized discrimination score was used as a measure to assess how well the forecasts are able to discriminate between varying observations. This was done by quantifying whether a set of observed outcomes can be correctly discriminated by the corresponding forecasts (Weigel and Mason 2011). The score measures the probability that any two (distinguishable) observations can be correctly discriminated by the corresponding forecasts. Thus, GDS can be interpreted as an indication of how often the forecasts are “correct” regardless of whether forecasts are binary, categorical, continuous, or probabilistic (Mason and Weigel 2009). The relative operating curve skill score was also used to compute the skills in tercile forecasts (i.e., probability forecasts for upper, middle, and lower terciles forecasts) considering rainfall forecasts only. The ROCSS measures the hit rate of a forecast against its false-alarm rate as the decision threshold (e.g., a quantile of a probabilistic forecast) is varied. It is expressed as a percentage and quantifies the improvement over climatological forecast (Jolliffe and Stephenson 2003). Characteristics of the relative operating curve (ROC) have been widely discussed (e.g., Kharin and Zwiers 2003; Mason 2003). Several other studies have used the technique to diagnose ensemble forecast accuracy (Ogutu et al. 2017; Gallus and Segal 2004; Legg and Mylne 2004).

Accessing, downloading, and analysis of data was carried out using relevant packages within R statistics: SpecsVerification (Siegert 2017), easyVerification (MeteoSwiss 2017), downscaleR (Bedia et al. 2017), visualizeR (Frías et al. 2017), and transformeR (Bedia and Iturbide 2017).

4. Results

a. Farmers’ expectations and hydroclimatic information needs for rice farming decision-making

In the face of difficulties posed by climate variability, farmers report having limited access to reliable sources of hydroclimatic information to support their farm decisions. Results showed that almost half of the farmers (43.5%) rely only on experiences and personal predictions based on indigenous ecological knowledge. For example, the croaking of a frog and the movement of ants from their hole is an indication that it will rain the next day. During the workshop, one farmer complained about existing hydroclimatic information available: “Those people [providers of the hydroclimatic information] are liars, I do my own thing and I don’t rely on them at all. When I say it will rain it will, except for a few occasions when it rains unexpectedly.” More than half of the farmers (56.5%) use indigenous forecasts alongside climate forecast information from GMET via radio or TV, and in some cases through ESOKO (an information service provider for agriculture) and from extension officers of the Ministry of Food and Agriculture. When asked about barriers for the use of hydroclimatic information, farmers mention inaccuracy and untimeliness of information, difficulties interpreting technical information, and language barriers. Another reason for farmers not to use forecasts provided by GMET is that these do not fit their purpose, and are provided at regional scale and do not match the situation in their communities. They showed good understanding of how hydroclimatic information could support their farm decisions and lives; frequently reported benefits include seed usage, rice yield, appropriate water management, saving money, and having enough food for family.

Throughout a farming cycle, farmers make decisions for which they require information on climate and water (Table 2). Preseason decisions require information mostly on rainfall onset, rainfall distribution, and rainfall amount. Decisions during the season such as land preparation and planting also require information on rainfall onset. Dam water level was highest on the priority list of farmers engaged in irrigation. Temporal distribution of rainfall is the most important information to determine when and how much fertilizer to apply and when to conduct pest and weed control. Wind speed and direction was most important for spraying weedicides. However, farmers expressed little need for this information as they already spray early mornings to avoid strong winds. Finally, rainfall cessation and temperature were the most needed information to start harvesting, although rainfall distribution and amount are critical to choosing a harvesting method. For instance, it is better to harvest with sickles and knives on wetter than normal fields than to use reapers or combine harvesters.

Table 2.

Percentage of farmers in need of particular information for farm. Boldface values represent the information needed by the highest percentage of farmers for a specific decision.

Percentage of farmers in need of particular information for farm. Boldface values represent the information needed by the highest percentage of farmers for a specific decision.
Percentage of farmers in need of particular information for farm. Boldface values represent the information needed by the highest percentage of farmers for a specific decision.

These results were further confirmed during the evaluation workshop. Farmers ranked rainfall distribution, temperature, and dam water level ranking as most important followed by total rainfall amount and onset as fairly important before cessation. Wind speed and direction were considered the least important among all the information needs. Temperature and precipitation patterns were found relevant by all farmers irrespective of geographical location or type of farming except for dam water level, which was top on the list of irrigating farmers (see Table S1). Farmers consider the timing of information provision as essential for making decisions and mobilizing resources for farming activities. When asked about which times they would prefer to receive seasonal climate and hydrological (dam water level) information, 74% preferred 1-month lead time, 24% preferred 2 months, and only 2% preferred 4 months lead time. Of the 42 rice farmers (IO and BIR) who required hydrological information, 67% preferred a month lead time and 33% preferred 2 months’ time lead.

b. Forecast evaluation

Following the needs assessment, the skill assessment of the ECMWF-S4 climate forecast was performed on three different lead times (i.e., 0, 1, and 2 months). Rainfall and minimum and maximum temperature were evaluated by validating forecast with observation to determine their performance in the study area.

1) Rainfall verification

Analysis of rainfall forecasts showed a general mixture of wet and dry biases (−2 to +1 mm day−1) (see Fig. S3). In most cases, however, rainfall was underestimated (dry bias) except for the upper west region where JJA (peak monsoon season) for all lead times and SON (monsoon cessation) (lead time 0) showed some spread of overestimation (wet bias) of rainfall which decreases with lead time. Dry bias was high in MAM (monsoon onset) (irrespective of lead time) compared to SON and then JJA. A wet bias was found largely in the western part of the northern region and upper west region for JJA for all lead times and in the upper west region only for SON (lead time 0). A change in bias with respect to forecast lead times could be attributed to the existing influence of local features such as surface topography [see also Ogutu et al. (2017)] and for that reason the initial conditions for which the model was run.

For rainfall there was a positive correlation (0.2 ≤ r ≤ 0.6) between forecasted rainfall and observations (significance correlation most grids) for the entire study area for all lead times of SON and JJA. MAM showed a mixture of lower positive and negative (−0.3 ≤ r ≤ 0.2) correlation for a large part of the northern region. Negative correlations were mostly found in the northern region at lead time 0 and 1 and the upper east and northern part of the western region at lead time 2. SON showed the strongest correlation followed by JJA then MAM. There is, however, no drastic change in correlation for each season per lead time, except in the northern region where MAM and SON showed low correlation at an increasing lead time (see Fig. S4).

Summarizing results showed that rainfall forecasts are able to discriminate between varying observations in large parts of the study area (Fig. 2). This is the case for all seasons except for the MAM period when the northern region (lead times 0 and 1 month) and upper east region (lead time 2 months) exhibited poorer skills. The SON rainfall forecast is found to be more skillful than that for JJA. MAM only showed patches of skill in the upper east and west regions (lead times 0 and 1 month) and for the northern region (lead time 2 months). Interestingly, while the skill of the forecasted rainfall generally decreases with lead time in JJA and SON, it improves slightly with lead time in MAM.

Fig. 2.

The generalized discrimination score for rainfall (JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

Fig. 2.

The generalized discrimination score for rainfall (JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

The year-to-year tercile performance of rainfall forecast for entire study area through the exploration of the observation position and the forecast probabilities are shown in Fig. 3. Tercile probability of 30%–100% dominated the entire 40 years period for all lead times. In general, SON showed higher skills than the other seasons, especially in the upper and lower tercile of lead time 0 and also at the lower tercile of lead times of 1 and 2 months. The skills within the upper tercile of SON reduce slightly with lead time while the rest differ with lead time. Below and above normal rainfall forecasts are generally more skillful than the climatological forecasts in all season, except in MAM (lead time 1 month) and JJA (lead time 0 and 1 month) where the lower and upper terciles showed poor skill. The lower tercile exhibited comparatively higher skill than the upper tercile in all seasons except MAM lead time 1.

Fig. 3.

Yearly tercile plot of forecast rainfall (1981–2010) over all of northern Ghana for MAM, JJA, and SON. (Shading shows tercile probability, white dots indicate the tercile of the observations for each particular year, and ROCSS is for the entire study area and for all years). Asterisks indicate significant score at 95% level.

Fig. 3.

Yearly tercile plot of forecast rainfall (1981–2010) over all of northern Ghana for MAM, JJA, and SON. (Shading shows tercile probability, white dots indicate the tercile of the observations for each particular year, and ROCSS is for the entire study area and for all years). Asterisks indicate significant score at 95% level.

2) Minimum temperature verification

Tmin forecasts showed a dominating cold bias (up to −2.8°C) for large areas and for all seasons irrespective of lead time (see Fig. S5). There were, however, spots of warm bias in the northeastern part of the northern region for JJA lead times of 1 and 2 months (stronger for 2 months than 1). MAM showed higher cold bias compared to SON and JJA. Each season showed similar trends of cold biases irrespective of lead time.

Despite the recorded biases, forecast and observed Tmin showed positive correlation (0.4 ≤ r ≤ 0.6 dominating grid cells) in MAM, JJA, and SON of almost all areas of the study and for all lead times. The correlation in MAM and SON forecasts is weaker in some grid cells but nearly constant in JJA with lead time. SON showed some patches of poor correlation in the northwestern and northeastern part of the northern region at a lead time of 2 months and the extreme eastern part of the upper west at lead times of 1 and 2 months (Fig. S6). A significant correlation was observed in most grid cells in JJA for all lead times. MAM also exhibited significant correlations in a large part of the study area except for the upper west region at a lead time of 2 months. Large areas of the northern region exhibited a significant correlation for lead time 1 than 2 before 0. The results of the generalized discrimination score in Fig. 4 showed considerable skill in the Tmin forecast in almost all the study areas and lead times. JJA is comparatively more skillful than MAM and then SON. Spots of poor skill in the western part of the upper west region and the southeastern corner of the northern region for all lead times were found in SON. The poor skill in these areas got poorer with lead time. Generally, the skill of the forecast (Tmin) was nearly constant with increasing lead time.

Fig. 4.

The generalized discrimination score for minimum temperature (Tmin; JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

Fig. 4.

The generalized discrimination score for minimum temperature (Tmin; JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

3) Maximum temperature verification

Tmax showed a cold bias in all parts of the study area for all seasons and lead times (Fig. S7). SON showed higher cold bias compared to JJA and MAM. For all the seasons, cold bias showed nearly a constant change in bias with lead time. In spite of the dominating cold biases across the study area (Fig. S7), Tmax showed skill across the study area for all seasons and lead times (Fig. 5). Tmax exhibited a strong relationship between its forecast and the observation in most of northern Ghana (Fig. S8). The relationship between Tmax forecast and observation was generally better in MAM compared to JJA and then SON. It was, however, comparatively weaker at a lead time of 1 month for JJA and lead times of 1 and 2 months for SON. The correlation therefore reduced with lead time for all seasons but not consistently. MAM and SON at lead time 0 showed a significant correlation in all parts of the study area. However, SON (lead times 1 and 2 months) showed no statistical significance correlation. Few grid cells in the northern region showed significant correlation for all lead times in JJA and for lead times of 1 and 2 months in MAM.

Fig. 5.

The generalized discrimination score for maximum temperature (Tmax; JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

Fig. 5.

The generalized discrimination score for maximum temperature (Tmax; JJA, MAM, and SON) ECMWF-S4 forecasts against verifying observations from WFDEI for 1981–2010.

Tmax showed extensive skills (MAM higher than JJA and then SON) across the study area for all seasons and lead times (Fig. 5). Nonetheless, spots of poorer skills were seen at southern parts of the northern region at lead time 1 month of SON and at the south and north edges of the northern region at lead time 2 months of SON. Tmax recorded slight decrease in skills over lead times for MAM and SON while JJA recorded a reduced skill from lead time 0 to 1 month but an increase at a lead time of 3 months.

4) Accuracy and association of forecast and verifying observation

For all the studied variables (rainfall and minimum and maximum temperature), bias is found in ECMWF-S4. The model showed a dry bias for rainfall (Prcp) and a cold bias for minimum temperature (Tmin) and maximum temperature (Tmax) in large areas of the study region. While the dry bias and cold bias dominate rainfall and temperatures simulations respectively, the wet bias in JJA rainfall is seen in the western part of the northern and upper west regions. The existence of bias in the forecast may be due to the inability to accurately simulate the mesoscale systems over West Africa (Afiesimama et al. 2006). Forecast lead time was observed to have little to no effect on the bias and in most times the change was not consistent. Unlike Tmin and Tmax, which showed similar bias in all seasons, rainfall exhibited a unique bias in each season. The reason could be differences in mechanisms associated with each season and variation in local features such as vegetation and topography (Indeje et al. 2000).

Despite the biases in the forecast, an overall strong correlation was found between the forecast and observation. The correlation, however, was poor for MAM Prcp in the northern region, and the northern areas of the upper east and upper west regions. Tmin recorded the strongest correlation in all lead times of MAM and JJA. Tmax, on the other hand, showed correlation in each season but slightly reduce inconsistently with lead time. Prcp, Tmin, and Tmax generally showed some significant correlation in parts of the study area.

5) General performance of the forecast over the study area

Using the generalized discrimination score (Weigel and Mason 2011), the forecast was able to discriminate between varying observations and thus skillful over large areas of northern Ghana. Forecasted SON rainfall was more skillful than JJA and MAM. Lower rainfall predictability skills found in MAM could be due to the inability of ECMWF-S4 to adequately capture local features and processes. The skills of the forecasted rainfall were not severely influenced by the lead time. The skills exhibited by both Tmin and Tmax were homogeneous. For Tmin, however, JJA exhibited slightly higher skill compared to MAM and SON. Tmax showed higher skill in MAM as compared to JJA and SON. Good skill in Prcp, Tmin, and Tmax for all seasons and lead times makes ECMWF-S4 seasonal climate forecast potentially able to meet the identified hydroclimatic information needs of farmers. A summary of the skill according to season and lead time is shown in Table S4.

5. Discussion

The main aim of this paper was to study rice farmers’ hydroclimatic information needs in northern Ghana and to assess the performance of the ECMWF-S4 seasonal climate forecast in meeting those needs. The study provides better insight for understanding a demand-driven climate information services; farmers have critical seasonal hydroclimatic information needs and unequivocally require these information within a particular time period for adaptive decision-making.

Results show that almost half of farmers rely on indigenous forecasts for their farm decision-making. These forecasts are based on observations matched with long time experiences. Also, Gwenzi et al. (2016), Zuma-Netshiukhwi et al. (2013), and Roncoli et al. (2002) observed that farmers highly depend on indigenous forecasts for most farm decision-making. Meanwhile, some other farmers use climate information from the national meteorological agency (GMET) and other private communication organizations such as ESOKO. This result is consistent with existing reports on climate information services in Ghana (Gbetibouo et al. 2017; ESOKO 2016; Nderitu and Ayamga 2013), which state that climate information services are gradually taking root in Ghana and have the potential to help farmers survive adverse effects of climate variability and change. However, almost all farmers find climate information from GMET and ESOKO unreliable and would therefore be willing to access and use improved climate information from alternative sources if it were available.

In an area with a constantly varying and changing climate, rice farmers could potentially improve their production if they have accessible and usable climate and water information. To do this, however, it is essential to have good inventory of key farming decisions that are responsive to climate and water information so that information generated forecast products can be tailored to support their farming decisions [see also Stone and Meinke (2006)]. Also, engaging farmers in formulating these needs will increase their trust for the forecast systems. We found that farmers’ information needs are linked with specific farming decisions and stages of the growing season, which makes the timing of providing information relevant. Key information needs relate to rainfall and temperature (cf. Iizumi and Ramankutty 2015; Lambert 2014).

Previous studies have shown how climate variability adversely affects yield and farmers’ decision-making and the challenges of providing accurate hydroclimatic information for adaptive decision-making vis-à-vis a seasonal time scale (Feleke 2015; Ouédraogo et al. 2015; Hansen et al. 2009). For seasonal forecasts, a lead time of a month and beyond has previously been a problem even for the best models, limiting their usefulness for farmers (Hansen 2002). The performance of ECMWF-S4 was mostly independent of season and lead times, which is promising for meeting farmers’ needs up to a lead time of 2 months.

This study uses seasonal average as a proxy to assess performance and discuss the possibility of meeting farmers’ needs. Further study is needed to make stronger claims on the predictability of each information need with ECMWF-S4. For example, onset and cessation are expressed in calendar dates while dam water level requires a hydrological method to determine its predictability. Nonetheless, the existence of skill in the analysis showed potential in predicting the identified information needs. For instance, skills in tercile predictability of above and below normal rainfall could provide information on rainfall amount and seasonal flow of water to the irrigation dam. Based on the results of the GDS and ROCSS analysis, Table S3 synthesizes these possibilities taking into account the limitations associated with the current analysis.

Generally, ECMWF-S4 is able to simulate well the interannual variability, spatial patterns, and structure of Prcp, Tmin, and Tmax for all seasons at different lead times except MAM in the northern region and the northern areas of the upper east and west regions (Figs. S4, S6, and S8). This has great implications since increasing rainfall variability results in higher risk for farmers (Ochieng et al. 2016; Graef and Haigis 2001). Rice farmers in northern Ghana already complain of loss of seeds at the beginning of the raining season due to delay in rainfall onset and variability between March and May (Ndamani and Watanabe 2014). While GDS and ROCSS are important attributes for assessing forecast skills, forecasts with high discriminative power may still be subject to systematic errors and may require postprocessing such as bias correction to become useful (Weigel and Mason 2011; Weigel et al. 2009). A bias of up to 2 mm day−1 and 2° to 3°C as observed in Tmin and Tmax could adversely affect farm decisions. These forecast biases could be attributed to the consequence of the intrinsic limitations of the physical models related to parameterizations, equation simplification, and uncertainties in the initialization procedure (Doblas-Reyes et al. 2013). Such biases could be mitigated through the application of bias correction techniques that are normally based on statistical methods using antecedent series of forecasts and observations (Peng et al. 2014; Piani et al. 2010; Weigel et al. 2009). However, studies have showed that bias-correcting a ECMWF-S4 probabilistic forecast does not necessarily improve forecast skill (Ogutu et al. 2017) but does enhance the usability of the forecast by improving the root-mean-square error (Barnston et al. 2015).

Finally, in this study we have used an interdisciplinary approach by combining a needs assessment with a forecast skill test, in order to assess the potential for meaningful climate services for local-level decision support. The approach enabled a broader contextualization of existing research on seasonal climate forecast verification and farmers’ information needs, which are often assessed in isolation. In this way, we are able to move away from a one-directional approach of looking at climate services to two directions, where needs and skills are clearly documented and synthesized. Our findings demonstrate the value of linking climate forecasts to farm-level decision-making. As such, this study contributes to the need of better matching hydroclimatic information services with needs of end users and important calls to improve climate services (Vogel et al. 2017; Street 2016; Reeves et al. 2015). Following Stone and Meinke (2006) we showed that developing appropriate interdisciplinary systems to connect forecast products with farm management is needed if uptake of weather and climate information by farmers is to be successful.

6. Conclusions

This paper has addressed key aspects of climate information services: matching information needs and forecast performance. Results shows homogeneity in rice farmers’ hydroclimatic information needs although some of these needs are ranked higher than others depending on the frequency of use and farming type. A majority of farmers prefer to receive hydroclimatic information within a month lead time for proper planning and decision-making. Our analysis concludes that this is possible. ECMWF-S4 possess some skill for forecasting Prcp, Tmin, and Tmax in northern Ghana. The skill varies per season and location but barely on forecast lead time, having significant implications for meeting rice farmers’ information needs in northern Ghana with improved seasonal climate forecast at different lead times. The ECMWF-S4 seasonal climate forecast therefore has potential to provide farmers with information that improves their farm decision-making. Yet, information services will require careful introduction to increase trust in using more tailored results from the forecast systems. Finally, we recommend that given the limitations of this study (discussed in section 3), further research is needed to make stronger claims especially on the predictability of each information need with ECMWF-S4.

Acknowledgments

This study is part of the EVOCA Project, financially supported by INREF–Wageningen University and Research, MDF West Africa, and other funding partners. We thank the farmers and management of Bontanga Irrigation Scheme for being part of the study as well the three anonymous reviewers for their overwhelming suggestions and insightful comments. We confirm that the authors have no conflict of interest.

REFERENCES

REFERENCES
Afiesimama
,
A. E.
,
J. S.
Pal
,
B. J.
Abiodun
,
W. J.
Gutowski
, and
A.
Adedoyin
,
2006
:
Simulation of West African monsoon using the RegCM3. Part I: Model validation and interannual variability
.
Theor. Appl. Climatol.
,
86
,
23
37
, https://doi.org/10.1007/s00704-005-0202-8.
Alves
,
O.
,
G.
Wang
,
A.
Zhong
,
N.
Smith
,
G.
Warren
,
A.
Marshall
,
F.
Tzeitkin
, and
A.
Schiller
,
2002
: POAMA: Bureau of Meteorology operational coupled model seasonal forecast system. ECMWF, 32 pp., https://www.ecmwf.int/sites/default/files/elibrary/2003/7694-poama-bureau-meteorology-coupled-model-seasonal-forecast-system.pdf.
Ambani
,
M.
, and
Fiona
,
P.
,
2014
. Facing uncertainty: The value of climate information for adaptation, risk reduction and resilience in Africa. 16 pp., https://careclimatechange.org/wp-content/uploads/2014/08/C_Comms_Brief.pdf.
Amekudzi
,
L.
,
E.
Yamba
,
K.
Preko
,
E.
Asare
,
J.
Aryee
,
M.
Baidu
, and
S.
Codjoe
,
2015
:
Variabilities in rainfall onset, cessation and length of rainy season for the various agroecological zones of Ghana
.
Climate
,
3
,
416
434
, https://doi.org/10.3390/cli3020416.
Amikuzuno
,
J.
, and
S. A.
Donkoh
,
2012
:
Climate variability and yields of major staple food crops in northern Ghana
.
Afr. Crop Sci. J.
,
20
,
349
360
.
Asante
,
F. A.
, and
F.
Amuakwa-Mensah
,
2015
:
Climate change and variability in Ghana: Stocktaking
.
Climate
,
3
,
78
99
, https://doi.org/10.3390/cli3010078.
Asare-Kyei
,
D. K.
,
J.
Kloos
, and
F. G.
Renaud
,
2015
:
Multi-scale participatory indicator development approaches for climate change risk assessment in West Africa
.
Int. J. Disaster Risk Reduct.
,
11
,
13
34
, https://doi.org/10.1016/j.ijdrr.2014.11.001.
Baltzer
,
K.
, and
H.
Hansen
,
2011
: Evaluation study: Agricultural input subsidies in sub-Saharan Africa. International Development Cooperation (DANIDA), 33 pp., https://www.oecd.org/derec/49231998.pdf.
Barnston
,
A. G.
,
S.
Li
,
S. J.
Mason
,
D. G.
DeWitt
,
L.
Goddard
, and
X.
Gong
,
2010
:
Verification of the first 11 years of IRI’s seasonal climate forecasts
.
J. Appl. Meteor. Climatol.
,
49
,
493
520
, https://doi.org/10.1175/2009JAMC2325.1.
Barnston
,
A. G.
,
M. K.
Tippett
,
H. M.
van den Dool
, and
D. A.
Unger
,
2015
:
Toward an improved multimodel ENSO prediction
.
J. Appl. Meteor. Climatol.
,
54
,
1579
1595
, https://doi.org/10.1175/JAMC-D-14-0188.1.
Bedia
,
J.
, and
M.
Iturbide
,
2017
: TransformeR: An R package for climate data manipulation and transformation. R Package version 0.0.14. https://github.com/SantanderMetGroup/transformeR/wiki.
Bedia
,
J.
,
M.
Iturbide
,
S.
Herrera
,
R.
Manzanas
, and
J.
Gutiérrez
,
2017
: DownscaleR: An R package for bias correction and statistical downscaling. R package version 2.0.0, https://github.com/SantanderMetGroup/downscaleR.
Benin
,
S.
,
A.
Nin Pratt
,
S.
Wood
, and
Z.
Guo
,
2011
: Trends and spatial patterns in agricultural productivity in Africa, 1961–2010. ReSAKSS Annual Trends and Outlook Report 2011. International Food Policy Research Institute (IFPRI), 92 pp., http://www.ifpri.org/publication/trends-and-spatial-patterns-agricultural-productivity-africa-1961-2010.
Bowen
,
G. A.
,
2009
:
Document analysis as a qualitative research method
.
Qual. Res. J.
,
9
,
27
40
, https://doi.org/10.3316/QRJ0902027.
Cofiño
,
A. S.
, and Coauthors
,
2018
:
The ECOMS User Data Gateway: Towards seasonal forecast data provision and research reproducibility in the era of climate services
.
Climate Serv.
,
9
,
33
43
, https://doi.org/10.1016/j.cliser.2017.07.001.
Crane
,
T. A.
,
C.
Roncoli
,
J.
Paz
,
N.
Breuer
,
K.
Broad
,
K. T.
Ingram
, and
G.
Hoogenboom
,
2010
:
Forecast skill and farmers’ skills: Seasonal climate forecasts and agricultural risk management in the southeastern United States
.
Wea. Climate Soc.
,
2
,
44
59
, https://doi.org/10.1175/2009WCAS1006.1.
Doblas-Reyes
,
F. J.
,
J.
García-Serrano
,
F.
Lienert
,
A. P.
Pintó Biescas
, and
L. R.
Rodrigues
,
2013
:
Seasonal climate predictability and forecasting: Status and prospects
.
Wiley Interdiscip. Rev.: Climate Change
,
4
,
245
268
, https://doi.org/10.1002/wcc.217.
Donkoh
,
S. A.
,
J. A.
Awuni
, and
R.
Namara
,
2010
:
Improving the efficiency of inland valley rice production in northern Ghana
.
J. Ghana Sci. Assoc.
,
12
(
2
), https://doi.org/10.4314/jgsa.v12i2.62818.
ESOKO
,
2016
: Access and use of seasonal forecast Information gives hope to farmers in northern Ghana. https://www.esoko.com/access-and-use-of-seasonal-forecast-information-gives-hope-to-farmers-in-northern-ghana/.
Feleke
,
H. G.
,
2015
:
Assessing weather forecasting needs of smallholder farmers for climate change adaptation in the Central Rift Valley of Ethiopia
.
J. Earth Sci. Climatic Change
,
6
,
312
, https://doi.org/10.4172/2157-7617.1000312.
Ferro
,
C. A. T.
,
D. S.
Richardson
, and
A. P.
Weigel
,
2008
:
On the effect of ensemble size on the discrete and continuous ranked probability scores
.
Meteor. Appl.
,
15
,
19
24
, https://doi.org/10.1002/met.45.
Frías
,
M. D.
,
J.
Fernandez
,
M.
Iturbide
, and
J.
Bedia
,
2017
: visualizeR: Visualizing and communicating uncertainty in seasonal climate prediction. R package version 0.2.1, https://github.com/SantanderMetGroup/visualizeR/wiki.
Gallus
,
W. A.
, Jr.
, and
M.
Segal
,
2004
:
Does increased predicted warm-season rainfall indicate enhanced likelihood of rain occurrence?
Wea. Forecasting
,
19
,
1127
1135
, https://doi.org/10.1175/820.1.
Gbetibouo
,
G.
,
C.
Hill
,
J.
Abazaami
,
A.
Mills
,
D.
Snyman
, and
O.
Huyser
,
2017
: Impact assessment on climate information services for community-based adaptation to climate change. Ghana Climate Services Report Ghana Country Report. CARE International ALP Programme, 50 pp., http://careclimatechange.org/wp-content/uploads/2017/07/Ghana-Climate-Services-Country-Report.pdf.
GFCS
,
2016
: Development and Delivery of Climate Services Research Dialogue 8, 19 May 2016. Global Framework for Climate Services, 12 pp, https://unfccc.int/files/science/workstreams/research/application/pdf/part2.1_wmo_dilley.pdf.
GIDA
,
2011
: Bontanga irrigation scheme. Ghana Irrigation Development Authority, Ministry of Food and Agriculture, accessed 11 September 2017, https://mofa.gov.gh/site/?page_id=3022.
GIDA
,
2016
: Irrigation schemes by regions in Ghana. Ghana Irrigation Development Authority, http://www.gida.gov.gh/schemes.php.
Graef
,
F.
, and
J.
Haigis
,
2001
:
Spatial and temporal rainfall variability in the Sahel and its effects on farmers’ management strategies
.
J. Arid Environ.
,
48
,
221
231
, https://doi.org/10.1006/jare.2000.0747.
GSS
,
2013
: 2010 population and housing census. National Analytical report. Ghana Statistical Service, http://www.statsghana.gov.gh/docfiles/publications/2010_PHC_National_Analytical_Report.pdf.
Gwenzi
,
J.
,
E.
Mashonjowa
,
P. L.
Mafongoya
,
D. T.
Rwasoka
, and
K.
Stigter
,
2016
:
The use of indigenous knowledge systems for short and long range rainfall prediction and farmers’ perceptions of science-based seasonal forecasts in Zimbabwe
.
Int. J. Climate Change Strategic Manage.
,
8
,
440
462
, https://doi.org/10.1108/IJCCSM-03-2015-0032.
Hansen
,
J. W.
,
2002
:
Applying seasonal climate prediction to agricultural production
.
Agric. Syst.
,
74
,
305
307
, https://doi.org/10.1016/S0308-521X(02)00042-2.
Hansen
,
J. W.
,
A.
Mishra
,
K. P. C.
Rao
,
M.
Indeje
, and
R. K.
Ngugi
,
2009
:
Potential value of GCM-based seasonal rainfall forecasts for maize management in semi-arid Kenya
.
Agric. Syst.
,
101
,
80
90
, https://doi.org/10.1016/j.agsy.2009.03.005.
IFAD
,
2012
: Ghana: Country programme evaluation. IFAD Publ. 84, International Fund for Agricultural Development, accessed 4 October 2017, http://www.ifad.org/evaluation/public_html/eksyst/doc/profile/pa/ghana2012.htm.
Iizumi
,
T.
, and
N.
Ramankutty
,
2015
:
How do weather and climate influence cropping area and intensity?
Global Food Secur.
,
4
,
46
50
, https://doi.org/10.1016/j.gfs.2014.11.003.
Indeje
,
M.
,
F. H. M.
Semazzi
, and
L. J.
Ogallo
,
2000
:
ENSO signals in East African rainfall seasons
.
Int. J. Climatol.
,
20
,
19
46
, https://doi.org/10.1002/(SICI)1097-0088(200001)20:1<19::AID-JOC449>3.0.CO;2-0.
Jolliffe
,
I. T.
, and
D. B.
Stephenson
,
2003
: Forecast Verification. A Practitioner’s Guide in Atmospheric Science. John Wiley & Sons, 240 pp.
Kanamitsu
,
M.
, and Coauthors
;
2002
:
NCEP dynamical seasonal forecast system 2000
.
Bull. Amer. Meteor. Soc.
,
83
,
1019
1037
, https://doi.org/10.1175/1520-0477(2002)083<1019:NDSFS>2.3.CO;2.
Kharin
,
V. V.
, and
F. W.
Zwiers
,
2003
:
On the ROC score of probability forecasts
.
J. Climate
,
16
,
4145
4150
, https://doi.org/10.1175/1520-0442(2003)016<4145:OTRSOP>2.0.CO;2.
Kranjac-Berisavljevic
,
G.
,
R. M.
Blench
, and
R.
Chapman
,
2003
: Rice production and livelihoods in Ghana: Multi-agency partnerships (MAPS) for technical change in West African agriculture. 86 pp., https://www.odi.org/sites/odi.org.uk/files/odi-assets/publications-opinion-files/3990.pdf.
Kumar
,
A.
,
A. G.
Barnston
, and
M. P.
Hoerling
,
2001
:
Seasonal predictions, probabilistic verifications, and ensemble size
.
J. Climate
,
14
,
1671
1676
, https://doi.org/10.1175/1520-0442(2001)014<1671:SPPVAE>2.0.CO;2.
Lambert
,
D. K.
,
2014
:
Historical impacts of precipitation and temperature on farm production in Kansas
.
J. Agric. Appl. Econ.
,
46
,
439
456
, https://ageconsearch.umn.edu/record/189144?ln=en.
Legg
,
T. P.
, and
K. R.
Mylne
,
2004
:
Early warnings of severe weather from ensemble forecast information
.
Wea. Forecasting
,
19
,
891
906
, https://doi.org/10.1175/1520-0434(2004)019<0891:EWOSWF>2.0.CO;2.
Liebe
,
J.
,
2002
: Estimation of water storage capacity and evaporation losses of small reservoirs in the Upper East Region of Ghana. Diploma thesis, Geographische Institute der Rhinischen Friedrich-Wilhelms-Universitat Bonn, 106 pp.
Mabe
,
F. N.
,
D. B.
Sarpong
, and
Y.
Osei-Asare
,
2012
:
Adaptive capacities of farmers to climate change adaptation strategies and their effects on rice production in the northern region of Ghana
.
Russ. J. Agric. Socio-Econ. Sci.
,
11
,
9
17
, https://doi.org/10.18551/rjoas.2012-11.02?nosfx=y.
Manzanas
,
R.
,
J. M.
Gutiérrez
,
J.
Fernández
,
M. D.
Frías
,
A. S.
Cofiño
,
E.
Sánchez
,
J.
Voces
, and
E.
Rodríguez
,
2012
: European provision of regional impact assessment on a seasonal-to-decadal timescale: Report on assessment and combination of S2D predictions, 26 pp., http://www.euporias.eu/system/files/D32.1_Final.pdf.
Manzanas
,
R.
,
L. K.
Amekudzi
,
K.
Preko
,
S.
Herrera
, and
J. M.
Gutiérrez
,
2014
:
Precipitation variability and trends in Ghana: An intercomparison of observational and reanalysis products
.
Climatic Change
,
124
,
805
819
, https://doi.org/10.1007/s10584-014-1100-9.
Mason
,
I. B.
,
2003
: Binary events. Forecast Verification: A Practitioner’s Guide in Atmospheric Science, I. T. Jolliffe and D. B. Stephenson, Eds., John Wiley, 37–76.
Mason
,
S. J.
, and
A. P.
Weigel
,
2009
:
A generic verification framework for administrative purposes
.
Mon. Wea. Rev.
,
137
,
331
349
, https://doi.org/10.1175/2008MWR2553.1.
Mason
,
S. J.
,
L.
Goddard
,
N. E.
Graham
,
E.
Yulaeva
,
L.
Sun
, and
P. A.
Arkin
,
1999
:
The IRI seasonal climate prediction system and the 1997/98 El Niño event
.
Bull. Amer. Meteor. Soc.
,
80
,
1853
1873
, https://doi.org/10.1175/1520-0477(1999)080<1853:TISCPS>2.0.CO;2.
Mdemu
,
M. V.
,
2008
: Water productivity in medium and small reservoirs in the Upper East Region (UER) of Ghana. Doctoral dissertation, Rheinischen Friedrich-Wilhelms-Universität, 159 pp.
MeteoSwiss
,
2017
: EasyVerification: Ensemble forecast verification for large data sets. R package version 0.4.2, https://CRAN.R-project.org/package=easyVerification.
Morris
,
M. L.
,
R.
Tripp
, and
A. A.
Dankyi
,
1999
: Adoption and impacts of improved maize production technology: A case study of the Ghana Grains Development Project. International Maize and Wheat Improvement Center (CIMMYT) Economic Program Paper 99-01, 38 pp., https://ageconsearch.umn.edu/bitstream/48767/2/ep99mo01.pdf.
Ndamani
,
F.
, and
T.
Watanabe
,
2014
: Rainfall variability and crop production in Northern Ghana: The case of Lawra district. 8 pp., http://hdl.handle.net/10173/1261.
Nderitu
,
M. J.
, and
T.
Ayamga
,
2013
: Making seasonal forecasts usable in Ghana and Kenya. 4 pp., https://careclimatechange.org/wp-content/uploads/2015/05/JotoAfrika12_FINAL.pdf.
Niang
,
I.
,
O. C.
Ruppel
,
M. A.
Abdrabo
,
A.
Essel
,
C.
Lennard
,
J.
Padgham
, and
P.
Urquhart
,
2014
: Africa. Climate Change 2014: Impacts, Adaptation, and Vulnerability, V. R. Barros et al., Eds., Cambridge University Press, 1199–1265.
Nin-Pratt
,
A.
,
M.
Johnson
,
E.
Magalhaes
,
L.
You
,
X.
Diao
, and
J.
Chamberlin
,
2011
: Yield gaps and potential agricultural growth in West and Central Africa. International Food Policy Research Institute, 140 pp., https://doi.org/10.2499/9780896291829.
Nkrumah
,
F.
,
N. A. B.
Klutse
,
D. C.
Adukpo
,
K.
Owusu
,
K. A.
Quagraine
,
A.
Owusu
, and
W.
Gutowski
Jr.
,
2014
:
Rainfall variability over Ghana: Model versus rain gauge observation
.
Int. J. Geosci.
,
5
,
673
683
, https://doi.org/10.4236/ijg.2014.57060.
Ochieng
,
J.
,
L.
Kirimi
, and
M.
Mathenge
,
2016
:
Effects of climate variability and change on agricultural production: The case of small scale farmers in Kenya
.
NJAS Wageningen J. Life Sci.
,
77
,
71
78
, https://doi.org/10.1016/j.njas.2016.03.005.
Ogutu
,
G. E.
,
W. H.
Franssen
,
I.
Supit
,
P.
Omondi
, and
R. W.
Hutjes
,
2017
:
Skill of ECMWF System-4 ensemble seasonal climate forecasts for East Africa
.
Int. J. Climatol.
,
37
,
2734
2756
, https://doi.org/10.1002/joc.4876.
Onyango
,
E.
,
S.
Ochieng
, and
A.
Awiti
,
2014
: Weather and climate information needs of small-scale farming and fishing communities in western Kenya for enhanced adaptive potential to climate change. Proc. 2012 JKUAT Mechanical Engineering Annual Conf. on Sustainable Research and Innovation, Vol. 4, Jomo Kenyatta University of Agriculture and Technology, 187–193, https://ecommons.aku.edu/eastafrica_eai/9/.
Ouédraogo
,
M.
,
R. B.
Zougmoré
,
S.
Barry
,
L.
Somé
, and
B.
Grégoire
,
2015
: The value and benefits of using seasonal climate forecasts in agriculture: Evidence from cowpea and sesame sectors in climate-smart villages of Burkina Faso. CCAFS Info Note 01-04, 3 pp., https://core.ac.uk/download/pdf/132678659.pdf.
Peng
,
Z.
,
Q. J.
Wang
,
J. C.
Bennett
,
A.
Schepen
,
F.
Pappenberger
,
P.
Pokhrel
, and
Z.
Wang
,
2014
:
Statistical calibration and bridging of ECMWF System4 outputs for forecasting seasonal precipitation over China
.
J. Geophys. Res. Atmos.
,
119
,
7116
7135
, https://doi.org/10.1002/2013JD021162.
Piani
,
C.
,
J. O.
Haerter
, and
E.
Coppola
,
2010
:
Statistical bias correction for daily precipitation in regional climate models over Europe
.
Theor. Appl. Climatol.
,
99
,
187
192
, https://doi.org/10.1007/s00704-009-0134-9.
QWeCI
,
2013
: ECMWF System 4 forecasts for malaria in Ghana. Quantifying Weather and Climate Impacts on Health in Developing Countries, 16 pp., accessed 11 August 2016, https://www.liverpool.ac.uk/media/livacuk/qweci/SYS4_Report_Ghana.pdf.
Reeves
,
J. L.
,
J. D.
Derner
,
M. A.
Sanderson
,
S. L.
Kronberg
,
J. R.
Hendrickson
,
L. T.
Vermeire
,
M. K.
Petersen
, and
J. G.
Irisarri
,
2015
:
Seasonal weather-related decision making for cattle production in the Northern Great Plains
.
Rangelands
,
37
,
119
124
, https://doi.org/10.1016/j.rala.2015.03.003.
Rockström
,
J.
, and Coauthors
,
2014
: Water Resilience for Human Prosperity. Cambridge University Press, 292 pp.
Roncoli
,
C.
,
K.
Ingram
, and
P.
Kirshen
,
2002
:
Reading the rains: Local knowledge and rainfall forecasting in Burkina Faso
.
Soc. Nat. Resour.
,
15
,
409
427
, https://doi.org/10.1080/08941920252866774.
Roncoli
,
C.
, and Coauthors
,
2009
:
From accessing to assessing forecasts: An end-to-end study of participatory climate forecast dissemination in Burkina Faso (West Africa)
.
Climatic Change
,
92
,
433
460
, https://doi.org/10.1007/s10584-008-9445-6.
Roudier
,
P.
,
B.
Muller
,
P.
d’Aquino
,
C.
Roncoli
,
M. A.
Soumare
,
L.
Batte
, and
B.
Sultan
,
2014
:
The role of climate forecasts in smallholder agriculture: Lessons from participatory research in two communities in Senegal
.
Climate Risk Manage.
,
2
,
42
55
, https://doi.org/10.1016/j.crm.2014.02.001.
Runge-Metzger
,
A.
, and
L.
Diehl
,
1993
: Farm household systems in northern Ghana. A case study in farming systems oriented research for the development of improved crop production systems. Nyankpala Agricultural Experiment Station Research Rep., 249 pp.
Salack
,
S.
,
B.
Sarr
,
S. K.
Sangare
,
M.
Ly
,
I. S.
Sanda
, and
H.
Kunstmann
,
2015
:
Crop-climate ensemble scenarios to improve risk assessment and resilience in the semi-arid regions of West Africa
.
Climate Res.
,
65
,
107
121
, https://doi.org/10.3354/cr01282.
Salack
,
S.
,
C.
Klein
,
A.
Giannini
,
B.
Sarr
,
O. N.
Worou
,
N.
Belko
,
J.
Bliefernicht
, and
H.
Kunstman
,
2016
:
Global warming induced hybrid rainy seasons in the Sahel
.
Environ. Res. Lett.
,
11
, 104008, https://doi.org/10.1088/1748-9326/11/10/104008.
SARI
,
2011
: Rice Sector Support Project. 2011 annual report, Savanna Agricultural Research Institute, 108–110, http://www.csir.org.gh/images/CSIR-SARI_Reports/CSIR-SARI%20Annual%20Report%202011.pdf.
Schlenker
,
W.
, and
D. B.
Lobell
,
2010
:
Robust negative impacts of climate change on African agriculture
.
Environ. Res. Lett.
,
5
,
014010
, https://doi.org/10.1088/1748-9326/5/1/014010.
Siegert
,
S.
,
2017
: SpecsVerification: Forecast verification routines for ensemble forecasts of weather and climate. R package version 0.5.2, https://cran.r-project.org/web/packages/SpecsVerification/index.html.
Stockdale
,
T. N.
,
D. L. T.
Anderson
,
J. O. S.
Alves
, and
M. A.
Balmaseda
,
1998
:
Global seasonal rainfall forecasts using a coupled ocean–atmosphere model
.
Nature
,
392
,
370
373
, https://doi.org/10.1038/32861.
Stone
,
R. C.
, and
H.
Meinke
,
2006
:
Weather, climate, and farmers: An overview
.
Meteor. Appl.
,
13
(
S1
),
7
20
, https://doi.org/10.1017/S1350482706002519.
Street
,
R. B.
,
2016
:
Towards a leading role on climate services in Europe: A research and innovation roadmap
.
Climate Serv.
,
1
,
2
5
, https://doi.org/10.1016/j.cliser.2015.12.001.
Sultan
,
B.
, and
S.
Janicot
,
2003
:
The West African monsoon dynamics. Part II: The “preonset” and “onset” of the summer monsoon
.
J. Climate
,
16
,
3407
3427
, https://doi.org/10.1175/1520-0442(2003)016<3407:TWAMDP>2.0.CO;2.
Trambauer
,
P.
,
M.
Werner
,
H. C.
Winsemius
,
S.
Maskey
,
E.
Dutra
, and
S.
Uhlenbrook
,
2015
:
Hydrological drought forecasting and skill assessment for the Limpopo River basin, southern Africa
.
Hydrol. Earth Syst. Sci.
,
19
,
1695
1711
, https://doi.org/10.5194/hess-19-1695-2015.
Vogel
,
J.
,
D.
Letson
, and
C.
Herrick
,
2017
:
A framework for climate services evaluation and its application to the Caribbean Agrometeorological Initiative
.
Climate Serv.
,
6
,
65
76
, https://doi.org/10.1016/j.cliser.2017.07.003.
Weedon
,
G. P.
,
G.
Balsamo
,
N.
Bellouin
,
S.
Gomes
,
M. J.
Best
, and
P.
Viterbo
,
2014
:
The WFDEI meteorological forcing data set: WATCH Forcing Data methodology applied to ERA-Interim reanalysis data
.
Water Resour. Res.
,
50
,
7505
7514
, https://doi.org/10.1002/2014WR015638.
Weigel
,
A. P.
, and
S. J.
Mason
,
2011
:
The generalized discrimination score for ensemble forecasts
.
Mon. Wea. Rev.
,
139
,
3069
3074
, https://doi.org/10.1175/MWR-D-10-05069.1.
Weigel
,
A. P.
,
M. A.
Liniger
, and
C.
Appenzeller
,
2007
:
The discrete Brier and ranked probability skill scores
.
Mon. Wea. Rev.
,
135
,
118
124
, https://doi.org/10.1175/MWR3280.1.
Weigel
,
A. P.
,
M. A.
Liniger
, and
C.
Appenzeller
,
2009
:
Seasonal ensemble forecasts: Are recalibrated single models better than multimodels?
Mon. Wea. Rev.
,
137
,
1460
1479
, https://doi.org/10.1175/2008MWR2773.1.
Willmott
,
C. J.
,
S. M.
Robeson
, and
K.
Matsuura
,
2012
:
A refined index of model performance
.
Int. J. Climatol.
,
32
,
2088
2094
, https://doi.org/10.1002/joc.2419.
Zuma-Netshiukhwi
,
G.
,
K.
Stigter
, and
S.
Walker
,
2013
:
Use of traditional weather/climate knowledge by farmers in the South-Western Free State of South Africa: Agrometeorological learning by scientists
.
Atmosphere
,
4
,
383
410
, https://doi.org/10.3390/atmos4040383.

Footnotes

Supplemental information related to this paper is available at the Journals Online website: https://doi.org/10.1175/WCAS-D-17-0137.s1.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Supplemental Material