PREDICTION OF DRY SPELLS DURING GROWING SEASON WITH RESPECT TO MAIZE CROP IN NIGERIA

Abstract

The prediction of dry spells during growing season with respect to maize crop was carried out in nine stations which include Calabar, Warri, Ibadan, Ilorin, Lokoja, Makurdi, Yelwa, Kaduna and Yola in Nigeria. The main data, which span 1971 through 2013, used for this work were: daily rainfall, maximum and minimum temperature, 0600 and 1500 GMT relative humidity, wind speed at 2 metre level and sunshine hours. They were sourced from the Nigerian Meteorological Agency (NiMet), Lagos. The same data set (excluding rainfall and sunshine hours) from 0.125° ERA INTERIM Reanalysis, 1979-2013, and daily 0.25° horizontal resolution 3B42 rainfall from Tropical Rainfall Measuring Mission (1998-2013) were obtained to serve as supplement to NiMet data. The daily and average reference evapotranspiration (ETo) were computed for tne selected stations for maize crop 118 days gestation using the method described in FAO Irrigation and Drainage Publication 56 and Penman-Monteith combination equation. Dry spell frequencies were determined during maize growth. The percentage frequency of dry spell lengths categorized as < 5 days, 5-10 days, 11-15 days, 16-20 days and >20 days were determined on successive ten-day periods (dekads) from the onset of growing season to 120 days for the 43 years for maize crop . The mid-growing season stage critical dry spell onset dates, lengths and their trends were determined. Mann Kendall tests were performed on the trends of the onset dates and lengths of dry spell to ascertain their statistical significance. The initial and mid-season critical dry spell lengths in categories: <5, 5-10, 11-15, 16-20 and >20 days were analyzed in relation to maize yields. Yearly predictions of the onset dates and lengths of critical dry spells (predictands or classes) were made only during the mid-growing season stage with the use of Artificial Neural Network (ANN). The predictors (attributes) are mean values of maximum and minimum temperatures, mean temperature, 0600 and 1500 GMT relative humidity, wind speed at 2 metre above soil surface, vsunshine hours, net radiation and reference evapotranspiration. About 70% of the data set was deployed for training while 30% was for testing. Based on cross-correlation analysis which measured the relationship between the predictors (attributes) and the predictands (classes), seven different models were put forward for the prediction purpose. For the 9 out of the stations evaluated, the occurrence of more (less) critical dry spells of lengths 5-10 and 11-15 days during the mid-season and 5-10 days only during initial stages leading to less (more) maize yield was generally associated with El-Nino (La-Nina) years. The percentage frequency of mid-season spell lengths of category 5-10 days ranged from 4-31% for nine stations. The number of days maize farmers in all the stations could expect first and second mid-season critical dry spell occurrences after planting ranged from 35-82 and 50-86 days respectively. The most suitable model for the prediction of critical dry spell onset dates and lengths for the nine stations was Model 1 (9 parameters) followed by Models 2 (8 parameters), 3 (7 parameters) and 5 (5 parameters). Prediction lead times for first and second critical dry spell onset dates generally ranged from 2 weeks to 2 months in the nine stations. Assessment on the efficiency of the most suitable models (for onset dates and lengths) for the 9 stations based on statistics indicate that the root mean square error (RMSE), coefficient of determination (R2), Nash-Sutcliffe coefficient of efficiency (NSE), the Wilmott's Index of Agreement (WIA), RMSE-Observations Standard Deviation Ratio (RSR) and Prediction error margin ranged between 0.96 and 3.31; 0.58 and 0.93; 0.51 and 0.90; 0.82 and 0.98; 0.30 and 0.69; -4.56 to 4.89 days respectively. These results showed the capability of ANN to predict yearly onset dates and lengths of mid-growing season critical dry spells for maize crop. These findings will aid strategic and yearly planning of agricultural operations for enhanced maize crop yield in Nigeria.