Spatio-temporal variabilities of precipitation indices in Cameroon and neighbouring areas

Abstract

In this work, some climatic indices related to precipitation were calculated over Cameroon and some neighboring areas. Results underwent spatial and temporal analyses through appropriate mathematical tools. We used three datasets: precipitation data from 24 meteorological stations in Cameroon, simulated precipitation from four general circulation climate models (namely MPI-echam5, MRI-cgcm2.3.2a, BCCR-bcm2.0 and CSIRO-mk3.5) and CRU (Climatic Research Unit) gridded precipitation. These data cover different time periods from the range 1951-2005. Model data are also available for the last two decades of the 21st century, simulated under the SRES A2 emission scenario. The study of onset, retreat and length of the rainy season led to the grouping of stations into three different climatic zones and to the definition of an agricultural calendar. Results for other statistical parameters show that over the entire domain, precipitation frequency and intensity decreased while heavy rainfall increased. The two models CSIRO-mk3.5 and MPI-echam5 reproduce quite well observed patterns and gradients, however CSIRO-mk3.5 shows higher correlations. For the future period 2082-2098, onset dates are expected to be later by one pentad (5 days) or more and retreat dates later by less than half a pentad in most locations. This will lead to a slight decrease in the duration of the rainy season. The situation is reverse in Sahelian zone, where the season will be longer. Model projections for other statistical parameters show spatial variation in rainfall frequency and intensity except around Adamawa Plateau and in the Eastern part of domain. In the last part of the study aimed to lay a basis for the calculation of the Standardized precipitation index (SPI), data were firstly fitted to four probability density functions (i.e. gamma, exponential, weibull and lognormal). The appropriate distribution for each station was afterwards used for computation of the SPI. It appears that for short time scales (up to 6-month) and for stations above 10◦N, the gamma distribution is the most frequent choice, while below this belt, the weibull distribution predominates. For longer than 6-month time scales, there are no consistent patterns of fitted distributions and the number of stations showing bias to lognormal increases with the number of month. Results of SPI show that droughts were generally rarer than wetness. Long episodes of severe droughts and short episode of extreme droughts are identified in the decades 1970 and 1980 in many stations. Similar studies with CRU grid precipitation show good results with the best performances on short time scales. The study of operational drought thresholds shows the necessity of their objective definition because they change with space and with the considered time period. This will be useful in agriculture and water resource management