The Stochastic Hydrological Forecast Models for Rainfall Predictions to Aid Climate Change Resilience Within the Tana Alluvial Aquifer Zones Using the Arima Box-Jenkins Algorithms

Abstract

The amount of rainfall each year in an arid geographical setting makes the difference between life and death, to both humans and animal populations alike, in the area. Droughts and Flood events, variables that define climate change in Northern Kenya, are deemed stochastic, in that both possess Random Probability Distributions. To that effect, if a climate change variable like rainfall comes in excessively high levels above the normal, lives of both humans and animals may be lost. The Tana River County of Kenya and parts of the Garissa County are illfated to be located within such an un-envious geographical setting, experiencing both massive drought and flooding episodes in a climate change scenario of uncertainty. The loss extends to property as well, mainly those owned by the human populace living there. It borders Garissa County which bears similarly harsh climatic parameters. In cases of extreme drought (complete absence of rainfall) loss of lives are occasioned on the account of depleted pasture for animals and famine and/ or malnutrition for humans. To mitigate this extremity, stochastic and statistical models have been used to build a time series prediction model, which involves leveraging on the knowledge of rains information of today and yesterday, to predict the rains amount tomorrow, with a high precision. Secondary data of monthly rainfall in mm paraded over the past forty years were used to help make this model. The use of Autoregressive Integrated Moving Averages (ARIMA) and the Auto-Correlation Functions(ACF) correlograms were used to make the twelve-month rainfall time series predictions of Garissa and T/River counties within these parts that lie in proximity to the Tana Alluvial Aquifer's geographical reach, as well as the neighborhoods, to aid planning for intervention during extremity of either side (massive rains or prolonged drought in 2022) using the data of rainfall ranging from January 1981 to December 2021. The statistical concepts of Seasonality, Lags, and Differencing are well explained and the way they apply to developing and deploying the model to make the twelve month predictions, for the rainfall amounts expected in respective precipitation months. Used thus every year, these forecast monthly rainfall figures in the years to come and the respective temperature levels so predicted, will now inform the decisions undertaken by the national and local governments, as well as the Non-Governmental Organizations, NGOs, on how to have timely funding of flood or massive drought interventions, in the forecast months to be availed one whole year in advance. This will aid timely repairs for boreholes for drought preparedness, a timely deepening of wells within the shallow Tana Alluvial Aquifer, timely Earth-Dam and earth-pan repairs before the rains onset. Moreover, water tankering and the relocation of the persons settled in the Flood-risk prone riparian corridors of the Tana Flow courses will be undertaken in good time. The present study proves that the ARIMA Model is a useful stochastic planni...