Report to CFU on the analysis of rainfall data from Southern province in Zambia

Roger Stern, November 2004

Summary

Rainfall data are analysed for three stations in Southern Zambia. They showed no evidence of climate change to the extent that farmers should be changing their farming practices.

A variety of definitions of the start of the rains were considered that correspond to criteria that farmers used or that were given in the project proposal. An examination of the risk of a long dry spell after sowing indicates that sowing is safer with planting after mid-November, than earlier. But further delay does not reduce the risk.

If it is assumed that the flowering period for maize (the most sensitive period for drought) is of about 20 days duration, and this period should ideally start before the end of January.

A simple water balance index is used for the three stations to guide users on the risks of maize of different season-lengths and compared to millet.

1.  Introduction

This report is from the pilot contract “Using Climatological Data to reduce Crop Failure Risk for Small-Holders”. It covers the analysis of rainfall data from three stations in Southern Zambia.

Analyses are only as good as the underlying data. A separate report describes the computerised database that has been used, checked and supplemented through this project.

The evidence for climate change is one of the issues that led to this project. This is addressed in Section 4 of this report. It is considered in more detail in a case study that was stimulated by this project and was prepared in parallel with this work.

This report is designed to present the methodology, as well as to give the results. We anticipate that a later report will apply the same methods to at least 5 more stations in Southern province.

2.  Setting the scene

Conservation farming is not like a wonder drug that will cure all diseases. It is about adopting a set of farming practices that link together and are in harmony with your surroundings. Your surroundings consist essentially of the soil that you cultivate and the climate, particularly the rainfall, that permits the crops to grow.

The soil is within your control and many elements of conservation farming are designed so that you can continue to cultivate the same soil indefinitely. If you farm well, then the soil remains reasonably consistent from year-to-year.

The rainfall is outside your control and it varies considerably from year to year. The fact you cannot control the weather does not stop you from taking advantage of the benefits of the rainfall in good years, and trying to minimise the effects of a poor year. This report is designed to support you in making best use of the climate and hence also to avoid, as far as possible, the problems in a year when the rainfall is difficult.

The main rainfall months in Zambia are from October to May. The regions of Zambia view the rainfall differently. In the Northern provinces there is often too much rainfall, so flooding is a problem. In the South, which is the concern of this project, there is often not enough, and so the risks are concerned mainly with drought.

In this report we study drought partly through examining the risk of a long dry spell during the growing season. Many rainfall analyses have concerned themselves with monthly or decade totals, but to study dry spells we require access to the daily records. In Section 3 we show the type of data used for the analysis. A separate report describes how we obtained and checked the daily data for the stations analysed.

Everyone knows that there is “climate change”! There is worldwide publicity to this effect. In this report we are concerned with climate change in a specific way. Thus we want to assess the extent to which there is evidence for a change in the pattern of rainfall in Southern Zambia. If so, then to what extent should it affect farming practices. In Section 4 we look at the issue of climate change.

From Section 5 onwards we look at the date of the start of the season, at risks of dry spells, and then at the viability of different cropping practices.

Throughout this report we look initially at the data from Moorings (Monze). The results are then compared with Choma and Livingstone that are further to the South.

3.  The data

In Fig. 3.1 we show the eight stations that were analysed in this project. The data from Moorings, at Monze, provide an unbroken record of daily rainfall from 1922 to 2004, from the Tom Savoury farm. The daily records from Livingstone Met and from Choma Met from 1950 to 2004 are also analysed in this report.

Fig. 3.1 Southern Zambia with the rainfall stations marked

The type of data used for the analyses are shown in Fig. 3.2 and Fig. 3.3. The rainy season is usually from November to March, though there is occasionally rain in April and even May. In Fig. 3.2 we show the data from 1999 to 2000, which we call 1999 data here. This was a good year with over 1000mm and evenly spread. It was a la Nina year.

In Fig. 3.2 we have marked a possible planting date, of 21st November, using the criterion of planting as soon as possible after 15th November. In this year there were tempting rainfalls earlier, with a total of over 20mm on 9th and 10th November. Given the initial rain of over 50mm at the end of October, this was also a possible planting date, and the rains later in November indicate that it would probably have been successful.

In this year the worst problem until the end of the rains in late March is seen from Fig. 3.2 to be a 10-day dry spell in mid-December. But, with the heavy rains preceding that spell this would not have caused a problem for the crop. Hence a 125-day maize crop should have grown well in that year.

Fig. 3.2 Daily data, Moorings 1991 – annual total = 1033mm

The data from 1991, an El Nino year, provide a contrast. There were two occasions in mid-November that might have tempted planting, though neither reached 20mm. If planting had taken place one wonders whether maize would have survived the 13-day dry spell that followed. The 45mm on 8th December would have been used for planning (or replanting), but whether any crop would have survived the problems in January and February is doubtful.

Fig. 3.3 Daily data, Moorings 1991 – annual total = 395mm

The description above provides an indication of the analysis that is described in the following sections. After looking at the problem of possible climate change we will examine the start of the season and dry spells during the season. Then we look at the way in which the planting date can be used to plan the crop that has a reasonable chance of success.

4.  Climate change

We begin our study of climate change by examining the monthly rainfall totals. They are plotted in Fig. 4.1. Before reading onwards, we invite you to study Fig. 4.1 to assess for yourself the extent to which it indicates climate change, and if so, whether it should force a change in farming practice.

Fig. 4.1 Monthly rainfall totals from 1921 to2003

We consider the issue of climate change in more detail below, but our main conclusions follow from a visual inspection of the data in Fig. 4.1. They are as follows:

·  For Southern Zambia there is no evidence, from the rainfall data, of climate change to the extent that farmers should be changing their farming practices. They may have to change their practices for other reasons, like poor soil, but not particularly because of a change in the pattern of rainfall.

·  This lack of major change also permits us to use the long records of data that Zambia is blessed with, to estimate the comparative success of different cropping strategies.

·  Climate change can either be a change in the average, or a change in the variability. An increase in variability would imply that farmers now have more extremes to contend with, than was the case previously. In Fig. 4.1 this would indicate that recent years are more “squiggly” than those previously. Our inspection of Fig. 4.1 does not seem to indicate this.

In Fig. 4.2 we have added fitted curves to the data shown in Fig. 4.1. They indicate we may have been hasty, if a quick look at Fig. 4.1 was used to give the conclusion that there is no evidence for climate change. First we need to qualify the importance of the curves. Despite the long record, only one of these curves achieves “statistical significance” at 5% and that is the curve for December. We examine this curve in more detail. It explains just 8% of the variability in the monthly totals. The fitted curve (i.e. the estimated mean for December) rises from 180mm in 1922 to a peak of 224mm in 1959. Then it drops to an estimated 152mm in 2003.

Fig. 4.2 Monthly rainfall totals from 1921 to2003 with fitted curves

What is then apparent is that the “shape” of this curve in December, in Fig. 4.2, is similar to that of January and February. For example in January the estimated mean was 181mm in 1922. It rose to an estimated 243mm in 1946 and dropped back to 181mm in 2003.

We therefore looked more carefully at the data. The rainfall amounts on each rainy day are very variable, and we found no evidence for a change in the mean rain per rainy day. Hence the “signal” may be clearer if we look at the number of rainy days in the month and not just at the rainfall totals. We defined a rain day as one with more than 0.85mm. This seemingly arbitrary value is to avoid any complications at sites that may be inconsistent in their recording of very small rainfalls, and also to avoid possible complications in the use of inches and mm in the recordings.

The observed and fitted values for the number of rainy days are shown in Fig. 4.3, and do present perhaps a slightly clearer picture. For example, in December the fitted curve corresponds to an estimated average of 14 rain days in 1922, rising to 16 in 1959 and dropping to an estimated mean of 11 rain days in 2003.

Fig. 4.3 Number of raindays from 1921 to2003 with fitted curves

So, our conclusion now is that there is perhaps continual climate change. However, Moorings usually does have enough rain in the main months of December, January and February. The results from Fig. 4.2 and 4.3 are not so clear for November and March, and these are perhaps the key months in determining the cropping strategy.

For Moorings, we have proceeded as far as we can with the monthly data and hence look at some results from the daily values. In Fig. 4.4 we show the date of the start of the rains for one definition that we will use in Section 5. The definition is “the first occasion, from 15th November that has at least 20mm within a 3-day period, and then does not have a dry spell of more than 12 days in the subsequent 30 days”. With the inclusion of the dry-spell condition this might be termed a “successful planting date”.

The horizontal lines in Fig. 4.4 indicate the end of November and the end of December. With this definition we were able to plant successfully by 1st December in 49 of the 82 years. These data give no evidence of a trend in the pattern of planting during the period of the record.

Fig 4.4 Start date, Moorings 1922-2003
(First day with 20mm in 3 days from 15 November and no dry spell exceeding 12 days in next 30 days)

In Fig. 4.5 we show the length of the longest dry spell at any time in the 2-month period from 1st January. We see that most years have a dry spell of more than 5 consecutive days within this period and 6 of the years have a dry spell that exceeds 15 consecutive days. But again, we do not see any evidence that the pattern of dry spells is changing.

Fig. 4.5 Length of longest dry spell in Moorings, in January or February

In subsequent sections we look at other aspects of the cropping strategy. Our conclusion is that growing maize at Moorings is a risky strategy, but we do not see that climate change has affected the level of risks to any marked effect.

To conclude this section we look briefly at the data for Livingstone and Choma.

In Fig. 4.6 and 4.7 we show the same information as in Fig. 4.2 and 4.3, for Livingstone. These data are from 1950 to 2003, the period when the pattern in moorings showed a decline in December to February.

Fig. 4.6 Monthly rainfall totals for Livingstone from 1950 to2003

In Livingstone the pattern for December is similar, but January and February are not so clear, either for the rainfall totals, or for the number of rain days.

Fig. 4.7 Number of raindays for Livingstone from 1950 to 2003
Fig. 4.8 Monthly rainfall totals for Choma from 1950 to2003
Fig. 4.9 Number of rain days for Choma from 1950 to2003

There seems no evidence of a decline in the Choma data where the record is also from 1950 to 2003. The results are in Fig 4.8 and Fig 4.9.

Finally, to those who continue with the opinion that climate change has affected the variability, more than the mean, we query – “the variability of what?” To support farmers in their cropping strategy it is insufficient to say “the variability of the climate”. Is it the variability of the planting date, of the total during the main months, or the length of the season? Variability by itself is not enough. We need to say what it applies to. We have high variability of many key aspects, like the date of planting given in Fig. 4.4 and of the monthly totals, given in many of the figures above. This confirms that farmers have a risky time in Southern Zambia. But we have yet to see the evidence from these data, that the rainfall has become more variable and hence has changed the level of risk.

5.  The start of the season

The terms of reference in this project specified two alternative definitions of the start of the season, as follows: