Climate variability and its impacts on the agriculture system is clearly evident in Ghana. Weather and seasonal climate forecast information service has been in operation for some time in the country. However, farmers generally do not find the information useful for their farm-level decision making. Forecast accuracy, untimeliness, and mismatch of forecast information and needs are often reported constraints for farmers to use weather and climate information. Consequently, the majority of farmers rely on their indigenous ecological knowledge to predict weather and seasonal climate patterns. At the same time, current weather and seasonal climate forecast information systems in Ghana face serious constraints in how they are used (if at all because of the one directional assumption behind its development; where only science produces new knowledge and makes it accessible for end-users with no or limited involvement of the end-users. In this context, this study addresses the central question: How can climate information services be improved through the coproduction of farmers and scientist? It aims at improving the reliability and acceptability of forecast information by integrating indigenous and scientific forecast. In this dissertation, I used a multi-method research approach, consisting of social participatory methods, mental modelling methods, forecast verification methods, and the principle of citizen science for data gathering and analysis. Initial diagnostics revealed certain issues that limit the uptake of climate information services in Northern Ghana: (1) the mismatch between forecast information provided and the farmers' information need (2) poor quality of forecast information, (3) the disconnect between forecast providers (researchers) and farmers, (4) management of unrealistic expectations of farmers. In response, I proposed a framework for second generation climate services that have the potential to facilitate co-production of relevant and accurate weather and seasonal climate forecast information and manages user expectation while strengthening the collaboration between information providers and users. Results of our analysis show that farmers’ information needs are linked to the type and timing of farm-level decision making. Also, model-based seasonal forecasts have the potential to provide relevant information at farmers most preferred lead time. Findings also show that in addition to historical rainfall patterns, farmers also use observational changes in certain indigenous ecological indicators to predict the coming season. In particular, there is a cognitive relationship between the observational changes and the predicted rainfall event. I observed that farmers’ indigenous forecasting skills and techniques are not intuitive but rationally developed and improve with age and experience. Results also show that farmers and Ghana Meteorological agency are on average able to accurately forecast one out of every three daily rainfall events. Similar results were obtained at the seasonal timescale. Furthermore, I recognized that forecast reliability and usefulness can be improved if indigenous forecast data are quantitatively collected and integrated with the scientific forecast using the proposed integrated probability forecast method. Finally, this dissertation contributes to the calls for a more integrated, co-learning, and co-production approach to climate services that move away from the current focus on science-driven and user-informed climate services. The approach developed in this dissertation is relevant for managing the impact of climate variability and change, particularly because it includes the knowledge of indigenous peoples which is often overlooked.