ECMWF supports flood disaster response in Peru

Fatima Pillosu, Umberto Modigliani, Linus Magnusson (all ECMWF), Marti Bonshoms Calvelo (SENAMHI, Peru), Luisa Sterponi (consultant for Peruvian Defence Ministry), Maria-Helena Ramos (Irstea, France), Patricio Valderrama (COEN, Peru)


From March 2017, ECMWF provided Peru with its forecast products for a limited period of time to help the country deal with the exceptionally heavy rainfall it experienced in the first few months of the year. As early as 3 February, the government declared a state of emergency in all coastal regions. The most affected areas were in the north (Tumbes, Lambayeque and Piura). In Piura, several records for daily precipitation were broken: on 3 March in El Partidor, 258.5 mm was recorded; 121.6 mm was measured on 21 March in San Miguel; and between February and March in the area of Morropon 150 mm was exceeded on three occasions. In this area, in the past similar amounts have only been recorded during exceptional El Niño events, such as those seen in 1983 and 1998.

Figure 1
Precipitation observations. Average daily precipitation during March 2017 according to observations received from SENAMHI.

The rainfall led to rising waters in coastal ravines. In more southern mountainous regions, this led to what is known in Peru as ‘huaicos’, which are a mixture of water, mud and rocks. Several rivers burst their banks causing flooding and damage to housing and infrastructure in urban and rural areas; the failure of drainage systems; and disruption of the electricity supply and sewage treatment plants. As of 31 March, the disaster had left 101 people dead, 353 injured and 19 missing, while more than 200,000 homes had been destroyed or had become uninhabitable (figures from COEN, Centro Operaciones de Emergencia Nacional of the Peruvian Ministry of Defence).

Seasonal forecast

This anomalous rainfall is believed to be connected to warm sea-surface temperatures along the coast (a phenomenon called El Niño Costero), which were probably caused by an equatorial Kelvin wave in the ocean. This feature propagated from the Western Pacific, where it was first observed in the autumn of 2016 as a positive sea-surface height anomaly. Probably as a result of capturing the Kelvin wave early on, ECMWF’s seasonal forecast was able to predict the anomalous rainfall along the equatorial coast of South America. The forecast from 1 November 2016 showed a wet anomaly over the region in the February to April average.

Figure 2
Seasonal forecast. Ensemble mean anomalies for precipitation in the period February–April 2017 in the ECMWF seasonal forecast from 1 November 2016.

ECMWF’s response

On 26 March, ECMWF received a request for rainfall forecasts from the environmental expert deployed from France (Institut national de recherche en sciences et technologies pour l'environnement et l'agriculture, Irstea) to Peru by the European Union Civil Protection Mechanism (EUCPM) through the European Civil Protection and Humanitarian Aid Operations (ECHO). Due to the exceptional circumstances, ECMWF agreed to provide its forecast products to the Peruvian Meteorological and Hydrological Service (SENAMHI) and COEN for a limited period of time, in accordance with our rules for the distribution of real-time data.

Access to all web products and ecCharts was granted and experts with previous knowledge of ECMWF products facilitated the uptake by local services. ECMWF established technical contacts with staff at SENAMHI and provided access to binary data in GRIB format in order to allow local services to process the information through their visualisation and impact models. Access to a new test product, Point-Rainfall, was also granted. It consists in statistical post-processing of ECMWF ensemble forecasts (ENS) to produce probabilistic rainfall forecasts for points. The idea is to provide better guidance in cases of localised extreme rainfall.

Figure 3
Probabilistic rainfall forecasts for points. El Partidor in the Piura region saw 258 mm of rain on 3 March 2017, most of which fell between 4 p.m. and 10 p.m. local time. The charts relate to ECMWF forecasts of 12-hour precipitation issued on 27 February 2017 at 00 UTC (t+114 to t+126). They represent the 98th percentile for total precipitation from the raw ensemble forecast (left) and from the Point-Rainfall product (right). The risk area for heavy rainfall was well identified in both forecasts even five days in advance, but the raw ensemble did not suggest the possibility of the observed amount of 258 mm. The Point-Rainfall product, on the other hand, suggested that there was a chance, albeit a small one, of such an event occurring.

Use of ECMWF products

ECMWF web products helped SENAMHI forecasters to issue warnings of heavy rainfall that was likely to cause new flooding or to exacerbate existing flooding. Special attention was also paid to events that could hinder rescue operations and/or endanger rescuers’ lives.

Fugure 4
Weather warning. Example of a meteorological warning map issued by SENAMHI covering the period from 3 to 9 March 2017. The risk levels range from 'one' (no special precautions necessary – white) to 'four' (be extremely cautious – red).

The binary data was used to produce extreme precipitation forecast maps. The daily total precipitation forecast from ECMWF’s high-resolution forecasts (HRES) was combined with percentile maps generated from SENAMHI’s climatological and hydrological observations (PISCO), which is a gridded database for daily precipitation. The percentile maps showed the areas where the daily accumulated total precipitation (from 12 to 12 UTC) exceeded the 90th, 95th and 99th percentiles of the local climatology. This made it possible to highlight the areas facing a high risk of heavy rainfall and thus to issue corresponding warnings to COEN, the authorities, members of the Peru Disaster Risk Management System, the media and public users.

At the same time, ECMWF products were also used by the scientific team at COEN, who monitored the emergencies triggered by the precipitation and issued alert reports every three hours. ECMWF products were used alongside satellite images and local reports to enhance the accuracy of daily local rainfall forecasts and to enable better warnings of extreme rainfall and river floods. The use of forecast products by emergency teams in the field and during the early recovery phase not only improved preparedness for high-impact events but also helped to devise better-informed response plans. The cooperation with ECMWF has led SENAMHI to evaluate the possibility of acquiring a full NMHS (national meteorological and hydrological service) non-commercial licence to continue to have access to the full range of ECMWF forecast products.

Figure 5
High-resolution forecasts. Meteorological warning map and Categorized Rain Maps (CRM) issued for the warning from 21 to 23 April 2017, using ECMWF HRES (forecast issued on 18 April 2017 at 12 UTC). These charts do not cover days on which the rainfall was at its heaviest. The heaviest rainfall occurred in March, but SENAMHI did not have access to ECMWF data for those days.