Assessing the Impact of Spatio-temporal Drought Regimes Using Timescale Standardized Precipitation Index in Malawi

This study assessed the temporal and spatial drought regimes in Kasungu Agricultural Development Division from 1977 to 2017. The previous studies in Malawi only used the Percentile Index such that other indices including the timescale standardized precipitation index were not considered for drought analysis. This study bridged this gap by using 3month and 6-month Standardized Precipitation Indices in Drought Monitoring and Prediction software. This study found that the rainfall patterns are highly variable with decreasing trends as determined by high coefficient of variation values (> 0.5). These variations have subjected the region to frequent mild and moderate meteorological droughts. Based on driest years, the respective maximum drought severities were 5.358 in 1994, 3.638 in 1999, 3.492 in 2005 and 3.112 in 2015. Results suggest that drought severity is highly correlated to the rainfall variations in the corresponding districts. This study recommends usage of timescale indices to aid drought monitoring.


INTRODUCTION
Malawi has been experiencing droughts which are increasing in frequency, duration and severity. Over the past 20 years, dry spells and seasonal droughts have been so common [1]. The World Bank report [2] emphasized that natural disasters especially droughts and flash floods have been so recurrent in Malawi and Mozambique. These have caused different havoc in agriculture production, water resources and the economy of the country [3].
Precipitation in Malawi is the commonly available data in most districts [4]. Most farmers in Malawi depend on rainfed agriculture. However, the spatial-temporal climate variability has had a substantial impact on the amount, timing and frequency of rainfall resulting in recurring droughts [5]. These variations normally have negative impacts on agriculture production. Kasungu Agricultural Development Division (ADD) lies on the Kasungu-Lilongwe plain which is an agro-based region in Malawi.
Although significant research has been done on impacts of droughts in Malawi, research on drought characterization is still limited for most regions in Malawi. Despite most areas being vulnerable to drought conditions, no study has ever been done to characterize droughts in Malawi. Droughts are categorized based on their duration and frequency [6]. Hayes et al [7] further observed that meteorological droughts arise from rainfall deficits both in temporal and spatial resolutions. Though droughts may be meteorological, agricultural or hydrological, they all root from rainfall variations [8]. Temporal rainfall deviations from the normal results in rainfall variations resulting in droughts [9] . All droughts are as a result of deviations from normal conditions be it precipitation, soil moisture, stream flow or groundwater or aquifer recharge [10]. Rainfall, stream flow, and soil moisture datasets are necessary to understand drought variations in a region [11]. Of these parameters, rainfall is a key parameter for monitoring meteorological droughts [12]. Long term monthly time series data is important in calculating the SPI for drought monitoring [13]. Rainfall patterns and variations are best determined when long term time series data is used [14]. Forsythe et al [15] further recommended that long term rainfall data is required for drought climate vulnerability studies. The study undertaken by Szalai [16] compared two drought indices in Hungary and concluded that the SPI is a powerful drought index to be used in drought monitoring studies and Bordi [17] added to the discussion and agreed that the SPI is the mostly accepted index used in drought studies. The SPI is a comprehensive and an influential rain-based index in monitoring droughts [18]. The SPI can be used at different timescales making it the most used and most powerful drought monitoring index [19]. The major advantage of this index is that it requires rainfall as the only input. The World Meteorological Organization [20] recommends this index for meteorological drought analysis. However, the major disadvantage of this index is that it does not factor in other parameters that influence drought characteristics especially severity such as temperature and evapotranspiration [21]. This index to be effective for different time scales, longterm data which fits a parametric distribution is a need. The SPI analyses droughts at various temporal scales to enhance drought classification [22]. This extends the finding of [23] that drought analysis at various spatio-temporal scenarios aids in proper drought classification and development of relevant drought mitigation measures. The ability of the SPI to evaluate drought scenarios at different timescales augments its robustness [24]. This index is best used for meteorological drought analysis [25]. The principle of SPI is to convert long term rainfall data to a gamma probability density function which is then transformed to a normal distribution with a mean of 0 and a standard deviation of 1.
Being first of its kind, this study therefore does not depart from this line and sought to characterize and assess the impact of spatial and temporal drought regimes in Kasungu ADD in Central Malawi using the SPI 3-months (SPI-3) and 6-months (SPI-6) time scales. This study further sought to determine drought properties at the specified timescales which can be used as a basis for planning water resources projects such as calculating crop water requirements and crop diversification for irrigation in the region to combat the effects of droughts at different timescales.

MATERIALS AND METHODS
In this study, meteorological (rainfall and temperature) monthly data for 40 years (1977-2017) was collected from the Department of Meteorological Services. This data was used to calculate drought indices. The Digital Elevation Model (DEM) was downloaded from http://srtm.csi.cgiar and was used to determine the elevation configuration for the region. Descriptive statistics and normality tests were carried out in Statistical Package for Social Scientists (SPSS) software. SPI-3 and SPI-6 drought indices were calculated using AgriMetSoft Drought Monitoring And Prediction (DMAP) software application.
DMAP is a comprehensive tool that calculates meteorological, agricultural and hydrological drought indices [26].
Mathematically, SPI is calculated as in equation (1): where is the seasonal rainfall at the ith station and jth observation, is its longterm seasonal mean and is its standard deviation. However, rainfall data is normally fitted by a Gamma distribution when considering SPI. Taking this probability density function of the Gamma distribution, SPI is calculated as: (2) where Γ is the Gamma function; x is the rainfall (mm); is the shape parameter ( > 0 and is the scale parameter ( > 0 and: where is the mean of rainfall.
The SPI values are classified as shown in Table 1 below. Moderately dry -1.5 to -1.99 Severely dry ≤ Extremely dry When a drought has been determined, the runs theory as shown in Figure 2 was used to identify four drought properties namely duration, severity, intensity, and frequency. Figure 2: Drought characteristics using the run theory for a given threshold level Duration of a drought event defines the number of months in drought conditions. Severity defines the definite value of the sum of the index during a drought event. In other words, it is the sum of the integral area below zero of each drought event.
The intensity of drought is given by: Frequency of a drought defines the number of drought events the occurred within a given period. Mathematically, Figure 3 shows the Digital Elevation Model (DEM) of Kasungu ADD. Highlands within the range of 1500 -1700 metres above mean sea level lie on the south western part (Mchinji Mountains) and south eastern part (Dowa Hills). Some parts of Mchinji district, some parts of Dowa district and some parts of Ntchisi are at higher elevations such that they receive better amount of rainfall enough for crop production. Notably, these areas are less vulnerable to drought risks. This notion agrees with [27] who stated that areas at lower elevations are more prone to drought risks. Severity Drought intensity Duration =

Number of drought events Drought frequency
Total number of years of study period = Table 3 shows the summary statistical analysis of rainfall data in Kasungu ADD. The mean annual rainfall varies from 781.1mm for Kasungu to 1093 mm for Ntchisi. The higher rainfall values recorded in Ntchisi, and Mchinji could be attributed to the geographical nature of the areas which are mountainous as determined from the DEM, are coupled with forest reserves and are connected with a rich network of tributaries of rivers which supply a significant amount of atmospheric moisture for rainfall. The lower rainfall values could be due to extensive land use changes in the western part of the ADD. Farming is a core economic activity in the area such that most people have opened up more lands through the cutting down of trees and vegetation, leading to higher levels of deforestation. This reduces evapotranspiration rate resulting into low rainfall regimes. The coefficient of variation in all the districts shows that the annual rainfall in the region is not dependable and is highly variable (CV > 0.5). This agrees with [28] who noted that the higher the CV value of a distribution, the higher the variability of that distribution. The high values of standard deviation also suggest annual rainfall fluctuations and high variability of rainfall in the region. This agrees with [29] who found out that rainfall with a high standard deviation is considered more unpredictable than rainfall with a low figure. The knowledge and utilization of temporal and spatial pattern of rainfall trends analysed in this study is a simple yet essential requirement for agricultural planning and  Figure 5 highlights the rainfall trend significance in the ADD. The trend patterns indicated that the rainfall is decreasing significantly (p < 0.05) in the central low lying areas of Kasungu ADD. This suggests that the area is prone to droughts. The eastern part however indicated significant increasing trends (p < 0.05). The areas that indicated moderate rainfall variability had insignificant increasing and decreasing trends. The knowledge of rainfall variability and trend significance in the ADD would enable farmers to plant crop varieties which are resistant to weather shocks and crops which have low water requirements. This knowledge would also help water planners in the agriculture sector and water supply to develop systematic disaster risk management strategies. Assessing the Impact of Spatio-Temporal Drought ...

Figure 3. Digital Elevation Model of Kasungu ADD
Year XXXX Volume X, Issue Y, pp xx-yy Page assigned by journalJournal of Sustainable Development of Energy, Water and Environment Systems   From these observations, it showed that meteorological droughts are frequently experienced in Kasungu and Dowa district and less frequent in Mchinji and Ntchisi districts. This could be due to the topography of the districts in that Mchinji and Ntchisi mostly receive a considerably higher amount of rainfall which provides enough moisture for crop growth. The droughts which were observed could be due to the temporal and spatial variation of rainfall within the pre-stated time scales.
The spatial distribution of drought severity conditions was assessed using the Kriging interpolation technique using the point data. Using Kriging interpolation method Assessing the Impact of Spatio-Temporal Drought ...  In 1999, Figure 9 shows that most parts of Kasungu district exhibited higher drought severities with an average of 3.01. Lower drought severity was noted in Mchinji district. The higher levels of drought severities noticed were highly correlated to the rainfall variations in the corresponding districts. These droughts with high severities resulting from temporal and spatial rainfall variations affect agriculture yields in the ADD where most smallholder farmers depend on the rainfall for their production. This substantiates the study of [30] that noted that droughts reduce crop yields in agro-based economies.   Figure 12 (a) illustrated that Kasungu experienced the highest number of severe droughts (5) while Ntchisi had the least (1). Mchinji had more moderate droughts with a frequency of 6 with Ntchisi recording the least (2). Mild drought conditions were frequent in all the four districts. Figure 12 (b) displayed similar patterns based on SPI-6. This could be due to high rainfall variations which reduced the contribution of water in the water balance of the ADD. Both timescales indicated that the region is prone to droughts calling for extensive water management strategies in the region. This could include irrigation water management and rainwater harvesting.

CONCLUSION
The study was aimed at characterizing and assessing droughts in the spatiotemporal regimes in Central Malawi using SPI-3 and SPI-6. With the help of DMAP software, the SPI-3 and SPI-6 drought indices revealed that Kasungu ADD has experienced more of moderate and mild droughts (meteorological). This suggested a higher impact on agricultural production and hydrology of the ADD. Kasungu district experienced more droughts followed by Dowa, Ntchisi and Mchinji. The DEM showed that Mchinji, Dowa and some parts of Ntchisi have highlands which contribute to rainfall in the region. However, these regions contribute less amount of soil water content as most of the rainfall in such area is converted to runoff and not infiltration. The results indicated that areas which are at lower elevation and receive less rainfall are more vulnerable to meteorological droughts. This study therefore, recommends that SPI-3 and SPI-6 drought indices should be used to aid drought monitoring in areas where rainfall data is the only data collected. However, in areas with a lot of observed data sets, a combination of drought indices should be utilized to increase the dependability of the results.