Projects per year
The agricultural sector, one of the main drivers of the Mozambican economy, is affected by several challenges related to climate variability and change. Especially rainfall variability is reflected in the productivity of (rain fed) agricultural systems. Rainfall variability, in terms of its onset, cessation, amount and distribution, and related drought or floods, can lead to either low crop yields or total crop failure across the country. Agriculture early warning monitoring systems make use of near-real time vegetation indicators and rainfall estimates derived from satellite sensors. The use of satellite based rainfall estimates is especially valuable in a country as Mozambique, where rainfall shows high temporal and spatial variability, and the network of ground stations is very low. However, the quality of these datasets needs to be assessed. This study focuses on the validation of dekadal FEWSNET Rainfall Estimate v2.0 (available 2001-now) and TARCAT v2.0 (available 1983-now). Cumulative 10-daily observations (1994-now) from twelve rain gauge stations, distributed over all regions, were used for validating satellite based estimations in the overlapping period (2001-2012). Methodologies include (1) quantitative pixel-by-pixel comparison statistics, which evaluate the performance of the satellite products in estimating the amount of the rainfall, (2) categorical validation statistics based on overall and per-station contingency tables, which are used to assess rain-detection capabilities, and (3) overall and per-station cumulative probability functions. The first results indicate that both the FEWSNET RFE and TARCAT datasets show underestimations of very high rainfall events. The TARCAT dataset corresponds less to the observed values, showing higher bias, higher errors and lower efficiency. FEWSNET RFE shows a higher probability of detection, lower false alarm ratio and higher critical success index. The cumulative density functions highlight overestimations of TARCAT of very low rainfall events, and underestimations of higher and extreme rainfall events. The better performance of FEWNET RFE is most likely linked to the use of Global Telecommunication System (GTS) station data, possibly including some of the stations used in this study, for removing bias after processing the satellite data. An independent station dataset will be gathered in order to be able to perform an independent validation.
|Title of host publication||Proceedings of Global Geospatial Conference 2013 (AfricaGIS 2013, GSDI 14)|
|Publication status||Published - 2013|
|Event||Global Geospatial Conference 2013 (AfricaGIS 2013, GSDI 14), Addis Ababa, Ethiopia - |
Duration: 4 Nov 2013 → 8 Nov 2013
|Conference||Global Geospatial Conference 2013 (AfricaGIS 2013, GSDI 14), Addis Ababa, Ethiopia|
|Period||4/11/13 → 8/11/13|