Forecasting dry spells in Malawi

Analysis last updated 2021-07-16

Background

This document summaries the data analyses that were performed to link the occurrence of a dry spell with various climatological indicators. This work was done for OCHA’s Anticipatory Action pilot in Malawi related to dry spells. It explores: the relations with climatic indicators (observational data), the skill of a seasonal and 15-day forecast to predict dry spells, and the ability to observe dry spells in almost real-time. All work presented here was done by the Centre for Humanitarian Data with indispensable help from technical partners. All the code is openly available here and includes more detail on the analyses presented in this document.

This work uses a list of historical dry spells and rainy seasons that was created as part of this project. The list of historical events can be downloaded on HDX. Further information and analyses of historical dry spells can be found here.

Definitions

  • A dry spell is defined as at least 14 consecutive days with less than 2mm cumulative precipitation. This definition was provided by the World Food Programme (WFP) who also shared impact survey data (MVAC).
  • The rainy season onset is defined as the first day of a period after 1 Nov with at least 40mm of rain over 10 days AND no 10 consecutive days with less than 2mm of total rain in the following 30 days (DCCMS, according to Kimaro and Sibande (2008)).
  • The end of the rainy season is defined as the first day of a 15-day period after 15 March with 25mm or less of rain (DCCMS, according to Kimaro and Sibande (2008)). The cessation date is the first day of this 15 day period.
  • We only included data within the rainy season, unless defined differently.
  • In this analysis we focus on the Southern region (adm1), as this is where most dry spells were observed.
  • CHIRPS is used as data source of observed precipitation.
  • For part of the analysis we only focus on the months of January and February as these months were selected by the Anticpatory Action team as months most vulnerable to dry spells. December was also identified as a vulnerable month however the data were very noisy due to the rainy season sometimes starting in December: low rainfall was sometimes due to a late rainy season onset, sometimes to a dry spell. Therefore it was decided not to include this month in the first version of the AA trigger.
  • It is very important to be aware of the low statistical significance of the results presented in this analysis. This is due to the fact that the choices made (definition of a dry spell, the areas of focus, and the months to examine) result in only a few occurrences of dry spells in the last 20 years.

Correlation of observed indicators and dry spells

Optimally the pre-alert of the trigger would be based on indicators with a long lead time, preferably 3 to 6 months. This would give the possibility of a wider range of anticipatory actions to be implemented by agencies and implementing partners.

There are no forecasts predicting the occurrence of a dry spell in Malawi. Long-range forecasts predict various indicators such as the ENSO state, seasonal precipitation, or monthly precipitation. Since those indicators can be predicted with long lead time and were identified by partners as likely to be predictive of dry spells, we explored whether they have a correlation with the occurrence of a dry spell.

This analysis is thus solely looking at observational data, not at forecasts. If the correlations between these observed data sources turned out to be significant, the next step would be to move to forecasts. (We assume that if there is no such correlation in the observed data, there won’t be any in the forecasted data.)

Previous work

The body of literature on the relation between long-term climate indicators and the occurrence of dry spells in Malawi is rather limited. We are aware of two articles that investigated the correlation of dry spells in Malawi with ENSO, and 3-monthly precipitation, as well as temperature, wind speed, and wind direction. Mittal et al. (2021) investigated the relation between total precipitation in a 3-month period and the occurrence of a prolonged dry spell in Malawi. This work defines a dry spell as 14 consecutive dry days, where a dry day is a day with <=2mm of precipitation. They found that there was only a weak correlation between the occurrence of a dry spell and the total 3-monthly precipitation, where the strength of the correlation heavily depended on the station, i.e. on the geographical location. Streefkerk (2020) looked at the correlation between 5-day long dry spells and the climate-meterological indicators of temperature, wind speed, wind direction and ENSO strength. It should be noted that by definition 5-day long dry spells is a more common phenomenon than 14-day long dry spells, and thus results might not be transferable. Streefkerk reported that the chosen indicators do have some predictive value for the occurrence of the 5-day dry spells, while these correlations heavily depended on the station, i.e. geographical location. Moreover, the analysis suggests that the while the ENSO phenomenon has predictive value for overall drought, it is less decisive for the occurrence of dry spells as these are more local events.

ENSO state

A commonly used long-term climate indicator is the El Niño Southern Oscillation (ENSO) state. This state is a global phenomenon that causes seasonal climatological fluctuations and is related to Sea Surface Temperatures (SSTs). More background on the ENSO phenomenon can be found here. Our analysis explores the relation with the observed ENSO state and the occurrence of a dry spell.

Several different indicators exist to measure the ENSO state, of which most are listed here. The two indicators most commonly used are the NINO3.4 index and ONI. They use slightly different sources of data and aggregation methodologies but show similar historical patterns. For our analysis ONI was chosen because their data is commonly used to keep track of the current ENSO state. Both indicators could be used and no large differences in results are expected.

Malawi is in a transition zone where the impact of the ENSO phenomenon is mainly observed in the Southern region. In this region El Niño generally marks dryer weather, while La Niña generally causes wetter weather. This is shown in the figure below, which is adapted from IRI by Ileen Streefkerk. The hypothesis would therefore be that the El Niño state increases the likelihood of a dry spell occurring.

Left: El Niño impacts on rainfall, and Right: La Niña impacts on rainfall

Left: El Niño impacts on rainfall, and Right: La Niña impacts on rainfall

Data source

More explanation on the significance of ONI can be found here. Openly available ONI data exists from 1950 to present. Warm periods are defined as an ONI larger or equal to 0.5, while cold periods are defined as an ONI smaller or equal to -0.5. If 5 or more consecutive overlapping periods reach the threshold for a warm period, these periods are defined as experiencing El Niño state. Similarly, 5 or more consecutive overlapping periods that reach the threshold for a cold period are defined as experiencing La Niña state. All other periods are defined as in a Neutral state.

Methodology

For each 3-month period an ENSO state was assigned to the ONI data based on the definitions detailed above. A 3-month period was marked as having experienced a dry spell if a dry spell started during the middle month of that 3-month period. I.e. if a dry spell started on 15-03-2010, the FMA 2010 season was indicated to have experienced a dry spell. Since our main period of interest is the rainy season, this is the focus of this analysis. The rainy season was defined quite broadly, including all seasons from September-October-November (SON) to March-April-May (MAM). This is based on FewsNet’s calendar. Lastly, since the main effect of the ENSO phenomenon is expected in the Southern region, we only included dry spells that occurred in this region (which is the largest fraction of all dry spells).

Analysis

Below the ONI values per rainy season are shown. A red bar indicates the occurrence of a dry spell. The x axis shows the 3-month period the ONI value belongs to, which is abbreviated to the first letter of each month. From this graph we can already see that not all dry spells occur during an extreme ONI state, and not all extreme ONI states co-occurred with a dry spell.

We now classify the ONI values into the three states: El Niño, La Niña and Neutral. We then analyze which percentage of the occurrences of each state co-occurred with a dry spell (first graph), and the division of the occurrences of the dry spells across the states (second graph). This could be translated to false alarms and rate of detection respectively.

From these graphs we conclude that:

  • Two thirds (67%) of Neutral ENSO states co-occurred with dry spells, while one third (33%) of Neutral ENSO states did not. This is the largest percentage of an ENSO state co-occuring with dry spells.
  • When dry spells occurred, more than half (56%) of them occurred during El Niño or La Niña ENSO. However, the single most common state during which dry spells occurred was Neutral.

Due to this division of dry spells across ENSO states, we conclude that an ENSO state is not a reliable indicator of the occurrence of dry spells. It is especially surprising that the Neutral state shows the most correspondence with dry spells, while from literature the Niño state is expected to show a higher correlation.

We also further analyzed if more extreme ONI values, instead of only the ENSO state, showed a higher correlation. That was not the case. Moreover, we analyzed the results per 3-month period but no large differences were seen here either.

Due to not seeing a relationship between observed ENSO state and dry spells, we didn’t move on to analyze the forecasts. For those interested in the forecasts, the most commonly used forecasts is produced by IRI, of which the historical data can also be downloaded.

Below-average seasonal precipitation

Another commonly used forecast with a long lead time are seasonal forecasts (covering a 3-month period). Seasonal precipitation forecasts are an often used product for informing predictions related to seasonal drought and, depending on the source, can provide 1 to 6 months’ lead time.

Seasonal precipitation forecasts exist in different formats. The most common format is that of a tercile-based forecasts,where the three terciles are referred to as below-average, normal, and above-average precipitation. These tercile-based forecasts report the probability for the precipitation to be in each tercile, per raster cell. See here for a clear resource on the usage if tercile-based forecasts. The Malawi Met services (DCCMS) provide their forecast in the tercile format, as well as global organizations such as IRI and [NMME](https://www.cpc.ncep.noaa.gov/products/NMME/prob/PROBprate.S.html.

While we investigated the relationship of dry spells with seasonal below-average precipitation, there is an important difference to note between the concept of below-average precipitation and dry spells. The amount of rainfall that is classified as “below-average” depends on the average rainfall for the given raster cell. By definition below-average precipitation occurs on average once every three years for each raster cell. The tercile is thus location dependent, whereas the definition of a dry spell is based on an absolute number of millimeters and thus doesn’t depend on the location. This means that a dry spell might occur more often than 1 in every 3 years in one location, while it might never occur in another.

Data source

CHIRPS was used to compute the occurrences of observed 3-month periods with below-average precipitation. CHIRPS was chosen because this is the same source that was used to detect observed dry spells and thus thereby we eliminate any weakening of relationship due to biases between different sources. See for more information about CHIRPS the documentation on defining observed dry spells.

Methodology

We worked with the monthly precipitation total as directly provided by CHC on their FTP server. We then compute the 3-monthly sum per raster cell from this. We then determine for each raster cell whether it experienced below-average precipitation. Whether a cell experienced below-average precipitation is a binary variable based on the observed totals and contrary to the probability value provided by forecasts.

Once we had the information whether a raster cell had experienced below-average precipitation during a given season, we aggregated this information to admin2 level. For the aggregation to admin2 level, we used a percentage-based approach. Since whether a raster cell had below-average precipitation is a binary variable, taking the mean would not be appropriate. We therefore classified an admin2 as having experienced below-average precipitation if at least 50% of the raster cells were labelled as having received below-average precipitation.

A 3-month period was assigned as having experienced a dry spell if a dry spell started during any of the 3 months in a given admin2. Thus this methodology is slightly different than the one used in the ENSO analysis, where a dry spell was only assigned to a season if it started during the middle month of that season.

Analysis

The confusion matrix below represents the occurrence of dry spells and below-average precipitation. As can be seen the co-occurrence of the two is not highly frequent. Only 55% (52/94) of seasons with a dry spell also had below-average precipitation. Moreover, in 89% of the 3-month periods with below-average precipitation, no dry spell occurred. Further analyses were done to determine if better correlations occur in certain admin2’s or during specific periods. That analysis didn’t show a significantly stronger signal.

Based on this confusion matrix and the other analyses, we concluded that the occurrence of below-average precipitation is not a reliable indicator for the likelihood of a dry spell occurring. We therefore didn’t move on to seasonal forecasts.

Total monthly precipitation

Besides the long-term tercile forecasts, some organizations also provide absolute forecasted rainfall (in mm) with several months leadtime. This rainfall is forecasted either as an expected amount per day or per month. The Copernicus Climate Change Services provide data from several organizations forecasting monthly total precipitation. From the daily projected amounts, one could in theory directly forecast the occurrence of a dry spell. However, since those forecasts have quite large uncertainty, it is very unlikely that the forecast would predict the occurrence of a dry spell by calendar day. Nevertheless, one can come up with other aggregated measures that might correlate with the occurrence of a dry spell. In this analysis we investigate the relationship between dry spells and observed total monthly precipitation.

Data source

CHIRPS was used for the same reasons it was used to compute seasonal precipitation.

Methodology

CHIRPS directly report observed monthly rainfall. This was aggregated from raster cell to admin1 by taking the mean value of all cells within the admin1. The reason we aggregated to admin1 and not admin2 was because of the high spatial uncertainty in forecasts. This means that these forecasts are meant to indicate general regional patterns, but are not able to predict whether the rain will fall in the one city or the next. Therefore admin1 is a more proper spatial area to analyze if observed rainfall is a decent indicator for dry spells.

To perform the analysis, we aggregated the dry spell data, which is at admin2, to admin1. We assigned an admin1 as experiencing a dry spell if at least 3 of its admin2’s experienced a dry spell during that time.

Moreover, we had to convert the daily dry spell data to monthly dry spell data, since we were working with monthly precipitation and thus need a classification of a dry spell per month to assess the correlation between monthly precipitation and dry spells. We assigned a month as experiencing a dry spell if at least 7 days of that month were part of a dry spell.

Lastly, this analysis was done at a later stage of the project. By then it was decided to focus only on the months of December, January, and February (December was dropped at a later stage) because dry spells occurring in these months cause the largest impact. In other words, we only looked at 3 out of 12 months in a year. We disregarded whether these months were within the rainy season, which is mainly relevant for December as the rainy season has typically begun by January and February.

Analysis

Before looking at the correlations, it is important to note that these methodological choices significantly decreased the occurrences of dry spells. This is caused by the fact that we only looked at a part of the country, aggregated to admin1, and focused on 3 months in the year. This left us with 3 dry spells in December, 1 in January, and 3 in February from 2000 till 2020. This is very little, and therefore it is important to note that the results presented below may not be generalisable or capture accurate trends.

Below is the distribution of monthly precipitation, with and without dry spells.

From this figure we can see that

  1. Months with a dry spell do on average have less precipitation than months during which no dry spell occurred
  2. Some months with no dry spell have precipitation in the same range as months with a dry spell
  3. The range of precipitation differs per month
  4. The separability of the occurrence (vs non-occurrence) of a dry spell differs per month

For December months, the distinction between months with and without dry spells is less clear. This is largely due to the fact that the rainy season often only starts in December. Therefore lower amounts of rainfall can be caused by the dry spells but also by dry days before the beginning of the rainy season. It was therefore decided to only focus on January and February.

We investigated different thresholds of monthly precipitation and the ability to classify dry spells based on these thresholds. Precipitation for January and February months ranged from 80 to 380 mm, with a mean of 225 mm. The figure below shows the misses and false alarms of detecting dry spells for different thresholds per month. We can see that the distinction is pretty good (recall that there are only 4 events of dry spells therefore it is unclear whether this pattern can be generalised.)

NOTE: false alarms are defined as False Positives / (False Positives + True Negatives) i.e. the percentage of times where monthly precipitation was below x mm but there was no dry spell over the number of times monthly precipitation was below x mm.

We combined months to determine what (single) threshold could be applied to both months in a trigger.

Based on that figure, we applied a threshold of 170mm since this is where the lines of misses and false alarms cross. We can see that all occurrences of dry spells had <=170 mm. However, 6 months with less <=170 mm rainfall didn’t co-occur with a dry spell, which would mean a false alarm rate of 60% (6/10).

Below is a heatmap showing the dates during which a dry spell and/or <=170 mm was observed.

Since observed monthly precipiation and dry spells showed an acceptably strong correlation, we decided to also analyze the skill of monthly forecasts and its ability to detect dry spells. This is explained in the next section.

Correlation of forecasts and dry spells

As described in the previous section, the only long-term indicator that showed a reasonable correlation with the occurrence of a dry spell was monthly precipitation. We therefore analyzed the skill of these forecasts. Moreover, we investigated a 15-day forecast that could more closely predict dry spells but at the cost of a much shorter lead time.

Previous work

We are not aware of any work attempting to forecast the onset of a dry spell or the likelihood of dry spells in Malawi. Moving away from Malawi, Gbangou et al. (2020) researched the predictability of dry spells in Ghana. They analyzed how well the forecasts can predict the number of dry spells within a season, where a dry spell is defined as at least 5 consecutive days with less than 1mm of rainfall per day. They futher attempted to forecast the length of the longest dry spell. For the forecasting, ECMWF’s seasonal forecast as well as a statistical model based on Sea Surface Temperatures (SST) was used. The authors showed that the skill of these two forecasts depended on the lead time and the geographical area within the country. The found correlations were generally weak. Interestingly, the forecasts were showed to be better at predicting extreme years, though the correlations still were not strong. The authors argue that the forecasts have better skill than guessing based on climatologies and thus can be used to inform actions. Similar work has been done by Surmaini et al. (2021), but focusing on Indonesia and using NOAA’s CFSv2 seasonal forecast model. They report slightly higher correlations, again also showing that these correlations heavily depend on the area of the country. Because Indonesia and Malawi having different climates, it is uncertain how transferable these results are.

Monthly precipitation

Data source

A few organizations publish forecasts on monthly precipitation with several months leadtime. For this analysis it was opted to use ECMWF’s seasonal forecast, as this is an often used and trusted source, and the data is openly available on the Copernicus Climate Change Services.
ECMWF releases a forecast each month on the 13th day of the month. A forecast includes projected total precipitation per month for 1 to 6 months ahead. Note on “lead times”: in ECMWF forecasts “1 month ahead” refers to the month the forecast was released i.e. the 1 month “leadtime” is published two weeks into that month. Therefore “lead time = 2” forecasts the month starting two weeks after their publication.

ECMWF’s forecast is a probabilistic forecast, meaning it consists of several members (=models) each with a projected monthly precipitation. This is what in this document is referred to as % of members, and can be interpreted as a probability of the event occurring.

Methodology

For the computation of observed dry spells and monthly precipitation the same method was used as in the analysis of observational data. I.e. the numbers were aggregated to admin1 level. Again we only looked at the Southern region and only the months of January and February.

ECMWF’s forecast is produced as a raster, with a low resolution of 1 degree. This raster was upsampled to have the same resolution as the CHIRPS data. The mean was taken of all the cells with their centre in the Southern region for each ensemble member separately.

Analysis

Two parameters had to be set for a trigger, namely the cap of forecasted monthly precipitation in mm, and the probability that the precipitation will be below that cap. One of these two numbers had to be set first. We chose to first set the probability and next, to determine the optimum cap. A probability of 50% was chosen as this is a clearly interpretable probability, and is relatively high.

NOTE: We also tested other probabilities, but this didn’t lead to better results.

The figure below shows the distribution of probabilities with and without a dry spell per leadtime. From this figure it can be seen that the distribution of the months with and without a dry spell is a lot less separable than we saw with observed data. Only the data with a leadtime of 1 month shows a high separability, but this month information is a lot less usable, since it is only released midway into the month it is predicting for.

The figure below shows the miss (=trigger unmet but there was a dry spell) and false alarm rates for different mm thresholds across all leadtimes. Based on this the threshold was set at the point where the false alarm and miss rates intersect across all leadtimes. This is at 210 mm. This is a different threshold than what we saw for the observational data, which was at 170 mm. We chose the 210 threshold as this showed the best balance between detection and false alarms, as shown in the graph below. We investigated using the 170 threshold on the forecasts as well, but this lead to a much lower detection rate.

NOTE: False alarm rate is defined as False Positives / (False Positives + True Negatives) i.e. the percentage of times the monthly precipitation was below x mm, but there was no dry spell over the number of times the monthly precipitation was below x mm

With the threshold at 210, we computed the confusion matrix per leadtime:

Taking into consideration logistical constraints for deploying an AAF, it was decided that the most suitable leadtimes were 2 and 4 months. A leadtime of 2 months means that a forecast covering January is published mid December. A leadtime of 4 months means that a forecast for January is published mid October.

For the 2 and 4 months leadtimes we computed a confusion matrix per month, to understand if there are large differences between the months. If we combined the numbers for January and February, we can see that with a leadtime of 4 months, 75% (3/4) of the dry spells are forecasted, but 86% (19/22) of the activations would be false alarms. With a leadtime of 2 months, 50% (2/4) of the dry spells would be forecasted, and 83% (11/13) of the activations would be false alarms. We can thus conclude that we can detect most of the dry spells but this comes at a very high false alarm rate. Most of these false alarms are expected to occur in February.

To understand the frequency of the threshold being met and/or occurrence of dry spells across time, we looked at the heatmaps.

From this analysis we concluded that the skill of the monthly forecast to detect dry spells is weak. With the current threshold of 210 we detect most of the dry spells, but this comes at a cost of a very high false alarm rate. Especially with a 4 months leadtime, this would lead to reaching the trigger 80% (16/20) of the years in February.

NOTE: We examined the effect of lowering the threshold to reduce the number of false alarms and activations (=trigger met). Reducing the threshold results in fewer false alarms but also a dignificantly lower detection rate.

15-day precipitation

The data

Several organizations produce rainfall forecasts with a 15-day lead time, but most of them are not openly available. For this analysis, we used CHIRPS-GEFS. This is a forecast produced by GEFS and bias-corrected to the CHIRPS data. This forecast was chosen because it is openly available, has a long historical record, is well-acknowledged, and is bias-corrected to the same data that was used to create a groundtruth dataset of observed dry spells. CHIRPS-GEFS is available as raster data at 0.05 resolution. A forecast is produced each day, and these forecasts are available from 2000 till present, with a data gap in 2020. Each forecast indicates the projected cumulative precipitation during the next 15 days per raster cell.

Methodology

Due to the shorter leadtime, we expected the forecast to have a higher spatial accuracy, i.e. for there to be less geographical uncertainty in the forecasts than when using seasonal forecasts. We therefore aggregated the forecast to admin2 instead of admin1, as this leads to more detailed information without making incorrect use of forecasts. The raster cell values of the CHIRPS-GEFS forecast were aggregated to admin2 by taking the mean value across all cells with their centre within the admin2.

Analysis

We first analyzed the general performance of CHIRPS-GEFS by computing a bias plot. This plot shows on the x-axis the observed precipitation over 15 days per admin2, retrieved from CHIRPS data, and on the y-axis the forecasted minus observed precipitation. If the forecast were perfect, all values in the graph would be on a horizontal line at [y = 0]. We can see that CHIRPS-GEFS has the tendency to overpredict low amounts of rainfall while underpredicting high amounts of rainfall. Since we are interested in extremely low amounts of rainfall, this fact can be problematic for the forecasting skill of dry spells.

Despite its bias, the forecast might be good enough to detect dry spells. This is not the case as can be seen from the figure below. The forecasted dry spells using the 15-day forecast very often don’t overlap with the observed dry spells and their timing does not align with those of the observed dry spells.

We define a dry spell as “detected” if any part of the observed dry spell overlaps with any part of the forecasted dry spell. That is a permissive definition.

We also defined a forecasted dry spell as a forecast projecting less than 2mm of cumulative rainfall over 15 days. This is a slightly stricter definition than that of the observed dry spells, which is less than 2mm cumulative rainfall over 14 days. The reason for this stricter definition is simply that the forecast only produces the rainfall over a 15-, not 14-day period. We also experimented with setting a higher threshold, partly to make up for this one day longer dry spell requirement.

The confusion matrix for this is shown below, from which it can be seen that the performance is quite poor. Only 10% (4/39) dry spells were detected. Moreover, there were many false alarms (n=99), which lead to a false alarm rate of 96%.

Due to the tendency of CHIRPS-GEFS to overpredict, we also tested various thresholds of the forecasted rainfall. Since the median of overprediction for 0-2mm of observed rainfall is around 25mm, it is expected that with a forecasted threshold of 25mm most dry spells would be detected. As can be seen in the confusion matrix this is indeed the case. However, this comes at a large drawback of many more false alarms (n=734).

From these results we concluded that CHIRPS-GEFS is not a suitable forecast to predict dry spells in Malawi.

Real-time observation of dry spells

We performed a comparison of two sources of observed precipitation. Throughout this project we have used CHIRPS as source of observed precipitation. While showing generally accurate results, one disadvantage of CHIRPS is that there is a long timelag in publishing the results of up to 1.5 months (i.e. the observations for November may be published as late as mid January). CHIRPS also has a preliminary product which is published 2 days after the end of a pentad (Dinku et al. (2018)). However, we prefer a shorter latency and therefore investigated ARC2 for its validity in detecting dry spells. ARC2 only has a time lag of a few days. A rough comparison is shown in the heatmap below. The two sources correspond to a large extent. They don’t fully overlap, but the severe dry spell events were detected by both sources. We therefore conclude that ARC2 would be a valid source to use for near real-time monitoring of dry spells.

References

Dinku, Tufa, Chris Funk, Pete Peterson, Ross Maidment, Tsegaye Tadesse, Hussein Gadain, and Pietro Ceccato. 2018. “Validation of the Chirps Satellite Rainfall Estimates over Eastern of Africa: Validation of the Chirps Satellite Rainfall Estimates.” Quarterly Journal of the Royal Meteorological Society 144 (April). https://doi.org/10.1002/qj.3244.

Gbangou, Talardia, Fulco Ludwig, Erik van Slobbe, Wouter Greuell, and Gordana Kranjac-Berisavljevic. 2020. “Rainfall and Dry Spell Occurrence in Ghana: Trends and Seasonal Predictions with a Dynamical and a Statistical Model.” Theoretical and Applied Climatology 141 (1): 371–87.

Kimaro, T. A., and H. Sibande. 2008. “Trends of Rainfall and Maize Productivity in Malawi.” Tanzania Journal of Engineering and Technology 31 (June): 72–82. https://doi.org/10.52339/tjet.v31i1.418.

Mittal, Neha, Edward Pope, Stephen Whitfield, James Bacon, Marta Bruno Soares, Andrew J Dougill, Marc van den Homberg, et al. 2021. “Co-Designing Indices for Tailored Seasonal Climate Forecasts in Malawi.” Frontiers in Climate 2: 30.

Streefkerk, Ileen. 2020. “Linking Drought Forecast Information to Smallholder Farmer’s Agricultural Strategies and Local Knowledge in Southern Malawi.”

Surmaini, E, E Susanti, MR Syahputra, FR Fajary, and others. 2021. “Use of the Dry-Spell Seasonal Forecast in Crop Management Decisions.” In IOP Conference Series: Earth and Environmental Science, 648:012092. 1. IOP Publishing.