| dc.description.abstract | The now more rampant and severe droughts have become synonymous with the SubSaharan
Africa where they are a major contributor to the acute food insecurity in the
Region. Though this is not different from other regions in the world, the uniqueness
of the problem in the Sub-Saharan Africa countries is the ineffectiveness of the
drought monitoring and predicting tools in use in these countries. Accurate and
reliable drought forecasts, when delivered in a timely fashion and in formats that are
comprehensible to the targeted users, are a precursor to successful drought mitigation
strategies. There is a link between weather monitoring and droughts; accurate weather
monitoring can detect droughts occurrence long before they strike. In Sub-Saharan
Africa, resource-challenged National Meteorological Services are tasked with this
responsibility. Although these Services use well-calibrated weather stations that meet
World Meteorological Organisation‘s standards, the high cost of acquiring the stations
allows only a sparse deployment.
Despite this challenge, these institutions continue to provide regular climate
forecasts especially in form of Seasonal Climate Forecasts. The utilisation of these
forecasts by the small-scale farmers whose crops/livestock depend solely on rainfall is
still very low; they instead continue to consult their Indigenous Knowledge Forecasts
for their cropping decisions. This is partly because the Seasonal Climate Forecasts are
too supply-driven, too ‗coarse‘ to have meaning at the local level and the
dissemination channels are ineffective. Why small-scale farmers? Economies of most
countries in the Sub-Saharan Africa are agri-based with over 70% of food being
produced by small-scale farmers practicing rain-fed agriculture. The latter in
extremely responsive to weather patterns and a good rain season translates to bumper
harvest and hunger and despair otherwise.
Though the robust Indigenous Knowledge Forecasts that these farmers have
relied on since time immemorial has always worked, there is evidence that the
knowledge is under serious threat from events such as climate change and
‗modernisation‘. Some of these threats can be countered by blending it with the
Seasonal Climate Forecasts. On the other hand, incorporating Indigenous Knowledge
Forecasts into the Seasonal Climate Forecasts will improve its relevance (both locally
and culturally) and acceptability and hence boost their utilisation among the smallscale
farmers.
The advantages of this mutual symbiosis relationship between the two
forecasting systems have been recognised and pursued in a few initiatives, but with
little success. The main challenge is the inability of these initiatives to scale-up
beyond a region/community and two, the lack of micro-level weather data to validate
the forecast outcomes. Information and Communication Technologies (ICTs) canaccelerate this integration; this is the focus of this research. The thesis describes a
novel drought monitoring and predicting solution that is designed to work within the
unique context of small-scale farmers in Sub-Saharan Africa. The research started off
by designing a unique integration framework that creates the much-needed bridge
(itiki) between Indigenous Knowledge Forecasts and Seasonal Climate Forecasts. The
Framework was then converted into a Drought Early Warning System prototype made
up of three components; (1) Drought Knowledge; (2) Drought Monitoring and
Prediction; and (3) Drought Dissemination and Communication. To achieve
sustainability, relevance and acceptability, indigenous knowledge was integrated in
each of the three components while mobile phones were used as both input and output
devices for the system. In order to facilitate collection and conservation of indigenous
knowledge on drought monitoring, an elaborate Android-based mobile application
was developed while text-to-speech and speech-to-text plug-ins were incorporated to
cater for semi-illiterate farmers. Wireless sensor-based weather meters were
acquired, calibrated against conventional weather stations and deployed as a
compliment to the weather stations. This proved the hypothesis that, when deployed
in hundreds, these sensors are capable of extending the weather network coverage to
enhance weather forecasting by downscaling the reading of weather parameters to
tens of meters.
Weather data is a ‗gold mine‘ for many sectors of an economy and to allow
public access to drought monitoring system data, a comprehensive web portal and an
SMS-based component were also implemented. In order to collect real data for the
indigenous drought forecast aspect, a case study of two communities in Kenya
(Mbeere and Abanyole) was carried out. On completion of the system prototype,
participants from the two communities evaluated it; based on content and format of
the integrated forecasts, 90% of respondents gave a score of ‘excellent‘.
The complexity of the resulting system was enormous and to ensure that the
above diverse parts worked together, artificial intelligence technologies were heavily
used in developing the system. Artificial Neural Networks were used to develop
forecast models whose accuracies ranged between 75 and 98% for lead-times of one
day to four years. Fuzzy logic was used to store and manipulate the holistic
indigenous knowledge while intelligent agents were used to integrate all the subsystems
into a single unit. After evaluating it using over fourty years of historical
weather data from Kenya, Effective Drought Index was adopted for drought
monitoring because of its ability to quantify and qualify drought in absolute terms. | en |
