A study done by Brown et al. (2010) explores the relationship between large-scale climate oscillations and land surface phenology metrics to determine influence of climate variability on the growing season. Spatial models were used to examine the distribution and interaction of these effects as determined by Normalized Differential Vegetation Index (NDVI) data. Specifically, 26 years of recorded data from North Atlantic Oscillation (NAO), Indian Ocean Dipole (IOD), Pacific Decadal Oscillation (PDO), and the Multivariate El Nino Southern Oscillation (ENSO) Index (MEI) were used to identify the most significant positive and negative correlations for the four climate indices in Eastern, Western, and Southern Africa. In the regions examined, the study found that the start of season (SOS) and cumulative NDVI of the growing season (cumNDVI) were significantly affected by variations in climate oscillations. — Anastasia Kostioukova
Brown, M., de Beurs, K., Vrieling, A., 2010. The response of African land surface phenology to large scale climate oscillations. Remote Sensing of Environment 114, 2286–2296.
Satellite remote sensing has become a primary input to monitor food production in Africa. To produce enough food to feed their families, hundreds of millions of Africans rely on sufficient rainfall and moderate temperatures. These variables are sensitive to climate change. Therefore, understanding which climate oscillations are most influential and affect variation in phenology metrics from one year to the next can improve seasonal analysis and agriculture planning across the continent. The objective is to provide evidence of whether climate variability captured in the four indices has had a significant impact on the vegetative productivity of Africa during the past quarter century. The use of satellite imagery provides a unique vantage point for observing seasonal dynamics of the landscape that have implications for global climate change issues.
Although five phenology metrics were calculated in the model, only SOS and NDVI were considered in the analysis. Further, African growing seasons do not consistently fall within one calendar year. Therefore, SOS and NDVI were determined by using two 1.5 year time periods— Cycle 1: October year 1– March year 3; and cycle 2: April year 2– October year 3. In order to reduce noise during correlation and increase the signal for each climate index, the study aggregated monthly climate indices into four seasons: December, January, February (DJF); March, April, May (MAM); June, July, August (JJA); and September, October, November (SON). Using spatial models, Brown et al. examined how many pixels behaved differently under a null hypothesis, with a significance of less than a 0.1 p-value. Rejection rates of the null hypothesis indicated links between climate oscillations and crop yields.
Brown et al.’s results were consistent with observed patterns of farming and climate variation. For example, results for cycle 2 capture the rainy seasons in which most Eastern African crops are grown. These are known as the Ethiopian ‘belg’ season, the Somalian ‘gu’ season, and Kenya’s ‘long rains’ season. Overall results show that East Africa’s SOS and cumNDVI are particularly sensitive to PDO variations in March–April–May. In Western Africa, the PDO during the September–November period dominates the correlation surface for cumNDVI. While the growing season in Southern African is sensitive to variations in the ENSO as well as NAO. The IOD was found to have a virtually no influence on phenology in all three regions. Brown et al. determined that further research on large-scale climate oscillation could provide forecast into future agricultural production in Africa.