Background

Local-scale early warnings for drought – can this increase community resili­ence?

Impacts of droughts on rural communities, including small-scale farmers, can consist of acute food shortage, livestock deaths due to depletion of pasture and water resources, as well as having health implications. There might also be secondary impacts such as school drop-outs and migration from rural to urban areas (Eriksen et al., 2005). Drought monitoring and fore­cast is therefore critical in order to provide early warnings that facilitate early action in miti­gation the impacts. In order not to only make short-term fix solutions to drought related problems, a strategy for decreasing vul­nerability and mitigating drought will also need to con­sider regional climate change projec­tions.

We hypothesize that an optimal local scale drought mitigation system need to include the following (i) communicated of early warnings both to local community stakeholders (possible via agricultural extension services), and to local and regional authorities dependent for drought mitigation actions; (ii) locally defined drought indi­ces, as well as local stakeholder reporting of early signs of initiation of drought to authorities. Also in this context agricultural extension services may play an important role; (iii) as a com­plement to local “soft” reporting, include on-site monitoring of development of droughts. Wireless Sensor Networks (WSN) is a promising way forward in this context; (iv) models used for seasonal forecast need to be downscaled and coupled to hydrological models that can provide local-scale drought assess­ments, in line with locally defined drought indices, and that can be validated against on-site monitoring of drought.

The project aims to, based on a pilot study in Mozambique (Limpopo river basin), test how such a system could be created in reality. Possibilities and constraints will be analyzed and recommendations for design of drought monitoring and forecasting for local-scale early warnings as a way to increase community resilience will be provided. To address these ob­jectives, the project team assembles a strong constellation of researchers in hydro-climato­logical modeling (P. Graham, L. Andersson), WSN (J. Wikner), model-assisted participation (L. Andersson, J. Wilk), and vulnerability assessments (J. Wilk).

Background

Farmers continually make a multitude of decisions based on available knowledge and oppor­tunities. When implemented, the difference between success and failure can depend on many factors including the timing of the new enterprise. Community access to EWS information must be made available in a context where it is possible for farmers to take heed and utilize the information to their advantage. Regional and local authority access to EWS is only useful if the information is relevant to local drought related problems (as perceived by local stake­holders) and if there exists an institutional framework that enables mitigation.

Seasonal forecasts esti­mate how rainfall and other weather variables may develop over time scales of one to several months or more. The El Niño Southern Oscilla­tion (ENSO) is known to be a key driver for African rainfall variability and dynamical pre­diction systems have be­come increasingly skilled in predicting this (e.g., Graham et al. 2005). Within WMO cooper­ation, a number of international forecasting centers now provide global seasonal forecasts as a regular part of their prediction products under the designation of Global Producing Centers for long-range forecasts. However, despite progress in prediction capability, practical use of fore­casts at the local and regional level is limited, due to factors such as poor timing, lack of local tailoring of relevant drought indices and local validation of information (Orindi et al. 2007).

As is suggested and described by, e.g., Masinde and Bagula (2010) wireless sensor networks (WSNs) are promising technologies to survey larger rural areas in order to get a spatially and temporal monitoring of drought indices such as soil moisture or water levels. Such infor­mation can then be used together with seasonal forecasts and local stakeholder’s observations to provide early warnings since they can provide continuous updated information of the spa­tial and temporal development of droughts in a local to regional scale. The sensor networks employ a large number of motes/modules/sensors, i.e., distributed nodes recording or report­ing the measured data. These motes are also able to perform wireless communication - either with each other or to a central office. In Dappin et al. (2009), for example, areas in the size of 10 000 km2 are monitored with sensors in order to detect and predict drought. The mentioned authors refer to already existing; successfully implemented systems deployed in for example Bangladesh, India and southern Africa. Typically, the modules should preferably be very cheap, not only in terms of hardware cost, but also in terms of calibration, maintenance, etc. With cheaper modules one can also distribute more of them in a certain area. This is advanta­geous in four ways: (1) better geographical resolution, it will enable the mod­ules to communi­cate shorter distances (wireless via radio) which in turn will (2) save power consumption, (3) makes it possible to trade number of modules against accuracy, and (4) in­crease the robust­ness of the network since more units could fail or run out of power, but the overall system would still be able to operate.

Pilot study area and collaborating partners

Work at the Council for Scientific and Industrial Research (CSIR) in South Africa includes advanced techniques for performing seasonal forecasting in the region. They currently pro­duce an ensemble of multi-model global seasonal forecasts that are downscaled for use at lo­cal scales, both with dynamical and statistical methods. We will work closely with CSIR to develop appropriate tools for incorporating their seasonal forecasts into both hydrological modeling and tailor-made analyses for local users. CSIR is also a partner in the recently initi­ated DEWFORA Project (Improved Drought Early Warning and FORecasting to strengthen preparedness and adaptation to droughts in Africa) financed by the EU. We are welcome as an unpaid partner of DEWFORA if our project gets financing from Sida. We will coordinate closely with DEWFORA to ensure that research ef­forts between these different efforts com­plement each other, where our specific contribution is our integrated local focus with a local community-based perspective. The seasonal forecasts will be input to a hydrological model to assess impacts on water resources and produce drought indices for both seasonal and long-term climate time scales. Close collaboration with University of KwaZulu-Natal (UKZN) was established under PAMO and is still ongoing through a Research Links grant. This will con­tinue here with the hydrological modeling components of the research. The multi-purpose ACRU Agrohydrological Model will be used, which was developed at UKZN specifically for conditions in Southern Africa and has been used for a wide variety of hydrological applica­tions in the region. The selection of the pilot study area (a catchment of approximately i50-350 km2) will be made in collaboration with DEWFORA. Through linkage with DEWFORA, results will be disseminated through Climate Outlooks Forums through the SADC-CM re­gional cli­mate centre. The modeling work will be focused on the Limpopo river basin, with a pilot study basin selected in collaboration with the Mozambique partner in DEWFORA (Univesidade Eudardo Mondlane). The regional modeling work will, however, be relevant for the whole Limpopo drainage basin that also covers South Africa and Botswana. We will build on experiences and contacts from our previous work in the Sida/UNDP Project “Climate change impacts on water resources in the Pungwe basin” (shared between Mozambique and Zimbabwe (Andersson et al. 2011), as well as on our experiences and contacts within the Sida project PAMO, with emphasis on collaboration with University of KwaZulu-Natal (UKZN) (Andersson et al. 2010), where we, in addition to collaboration within PAMO, have an ongo­ing cooperation through Research Links. The Limpopo basin is particu­larly drought prone and we will have a good contact network through our collaborating part­ner, CSAG at the Univer­sity of Cape Town (UCT), have good contacts through previous studies. The main applicant (L. Andersson) is secretary of the Swedish UNESCO committee IHP (International Hydro­logical Program) and if the project is financed, a workshop on the potentials for drought miti­gation through early warnings will be arranged in Southern Africa in collaboration between the Swedish IHP and regional IHP committees in Southern Africa to local, regional and na­tional authorities. The most important forms of dissemination on the local scale will be through the network of agricultural extension officers and farmer associa­tions.

Research Plan

Assessment of institutional framework and communication channels

An inventory and gap analysis will be made of institutional frameworks and drought mitiga­tion and adaptation practices that have an impact on drought mitigation in the pilot area. As­sessment of previous drought warning experiences among authorities, agricultural extension service and local communities will be made. Workshops will be arranged with (1:1) lo­cal/regional authorities and (1:2) agricultural extension officers and representatives of farm­ers’ organizations to address previous experiences of early warning systems, present and pre­ferred ways of communicating information. At the workshops, factors will be determined, ranked, evaluated and com­bined into better understand trade-offs based on assessments of how vulnerability has been altered with the provided EWS information. Based on the output from these workshops ways of communicating EWS information in the pilot project will be agreed upon.

Locally defined drought indices and “soft” observations (local knowledge)

A participatory modeling approach will be used to identify local/regional needs of information both at the community level and at the authority level. The methodology is a based on and will further de­velop the methodology used in the Sida sponsored PAMO project (Andersson et al. 2010). The pro­cess will determine how local level EWS can most effectively be transmitted and used. In addition to assessments of what information that is perceived as relevant to include in drought indices, ways of most effectively pre­senting information and possible communication channels will be assessed. Rele­vant actors in this participatory process include local authorities, water utilities, farmer organizations, extension services and dam operators. An initial dialogue between the participating stakeholders will determine the requested set of climate indices. This will take place during two workshops: (2:1) exten­sion service farm advisers and communities, with emphasis on small-scale farmers, and (2:2) lo­cal/regional authorities. Early warnings based on sea­sonal forecasts with information corresponding to the requested set of indices will then be produced (see below) and continuously communicated during one year in a form corresponding to the requests during the meetings. One year later a second set of workshops will be held to assess the usefulness of the early warnings (WS 3:1, 3:2). During WS 3:1, 3:2 also information of the possible implication of climate change will be pre­sented and discussed. In connection to WS 2:1 and 2:2 it will also be assessed if any “soft information” (e.g. from flora, fauna, or from observations – without monitoring or visual observations related to drought) is used by com­munities to monitor or predict drought. If so, this information will be collected. During WS 3:1, 3:2 it will be assessed if this information (since the previous workshops, one year earlier) have been used for any drought mitigation purpose among communities and authorities. In addition, we will assess if the suggested “soft”, “local knowledge” indicators correspondence to “scientific” predictions and the actual development of drought during the year between WS 2 and 3

Wireless sensor networks

Wireless sensors will be used as a way to provide spatially and dense geographic information about trends of water deficit. The sensors will indicate when specified thresholds of dryness are met at several locations, which will trigger the sensors to communicating to each other about the situation and send an alert for need of drought mitigation. In addition the sensors will provide information valuable for local calibration of the used hydrological model.

A wireless sensor network will be established for the pilot sub-catchments. The design of the network with regard to distribution of sensors and the type of information to be collected will be decided in connection to workshops. Technically the following research questions will be addressed: (1) Evaluate existing WSN systems and choose the most promising ones pro­vided the pilot area to be tested in. Important metrics and parameters are cost, reliability, mainte­nance, coverage, etc. For example, it would be a direct coupling between coverage (number of modules) and cost. We refer to this cost function as the trade-offs between the price, power, information throughput (speed), range (coverage), integrity (robustness), and life time. (2) Evaluate the design of sensors and how these can be connected to the modules. The most promising parameter to measure is in our view the soil moisture, but if requested by commu­nity stakeholders, additional sensors can be connected to measure other parameters at no or very low extra cost. The layout of the sensor network with regard to spatial coverage, design and what to monitor will be discussed during WS 2:1 and 2:2, (3) Investigate and un­derstand the implications on the sensor/module design given the measurements to be done. One key issue is the power consumption which should be minimized. In order to avoid (or at least minimize the need of) batteries, we will investigate different energy scavenging tech­niques. The control of the energy scavenging has to be reliable and “intelligent”.

Regional seasonal and decadal drought forecasts

Climatological EWS information from downscaled seasonal and decadal GCM’s will be cou­pled to models for assessments of local-scale drought related information including water resources, crop yields and environmental indicators (e.g. fire indices), based on predictions from CSIR. An important part of the process will be verification of the forecasts. Using hindcasting methods, detailed trials will be made over historical periods to verify skill in pre­dicting the seasonal climate, hydrological, and other indices, related to crop yields or ecosys­tem services (e.g. fire indices) predictions. Two time scales will be used for assessment of impacts to water resources. The primary focus will be on forecasting seasonal impacts as de­scribed above. However, a complementary focus will be to look farther into the future and assess expected overall impacts of climate change on water resources for a time horizon of 50-100 years. Projections of future climate will be based on the multiple regionally downscaled climate model results currently being produced within CORDEX (Coordinated Regional Downscaling Experiment) under the WCRP (World Climate Research Programme). The aim of CORDEX is to produce an ensemble of regionally detailed climate change projec­tions for use in assessments that will contribute to the next IPCC AR Report. Africa is a pri­ority area for this work and downscaled results will become available from the end of 2011.