Impacts of Satellite-Based Rainfall Products on Predicting Spatial Patterns of Rift Valley Fever Vectors

Clement Guilloteau Laboratoire d’Etudes en Géophysique et Océanographie Spatiale (CNRS, IRD, Université Toulouse III, CNES), Observatoire Midi-Pyrénées, Toulouse, France

Search for other papers by Clement Guilloteau in
Current site
Google Scholar
PubMed
Close
,
Marielle Gosset Géosciences Environnement Toulouse (CNRS, IRD, Université Toulouse III), Observatoire Midi-Pyrénées, Toulouse, France

Search for other papers by Marielle Gosset in
Current site
Google Scholar
PubMed
Close
,
Cecile Vignolles Centre National d’Etudes Spatiales, Toulouse, France

Search for other papers by Cecile Vignolles in
Current site
Google Scholar
PubMed
Close
,
Matias Alcoba Géosciences Environnement Toulouse (CNRS, IRD, Université Toulouse III), Observatoire Midi-Pyrénées, Toulouse, France

Search for other papers by Matias Alcoba in
Current site
Google Scholar
PubMed
Close
,
Yves M. Tourre Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York

Search for other papers by Yves M. Tourre in
Current site
Google Scholar
PubMed
Close
, and
Jean-Pierre Lacaux ** Laboratoire d’Aérologie, Observatoire Midi-Pyrénées, Toulouse, France

Search for other papers by Jean-Pierre Lacaux in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Spatiotemporal rainfall variability is a key parameter controlling the dynamics of mosquitoes/vector-borne diseases such as malaria, Rift Valley fever (RVF), or dengue. Impacts from rainfall heterogeneity at small scales (i.e., 1–10 km) on the risk of epidemics (i.e., host bite rate or number of bites per host and per night) must be thoroughly evaluated. A model with hydrological and entomological components for risk prediction of the RVF zoonosis is proposed. The model predicts the production of two mosquito species within a 45 km × 45 km area in the Ferlo region, Senegal. The three necessary steps include 1) best rainfall estimation on a small scale, 2) adequate forcing of a simple hydrological model leading to pond dynamics (ponds are the primary larvae breeding grounds), and 3) best estimate of mosquito life cycles obtained from the coupled entomological model. The sensitivity of the model to the spatiotemporal heterogeneity of rainfall is first tested using high-resolution rain fields from a weather radar. The need for high-resolution rain data is thus demonstrated. Several high-resolution satellite rainfall products are evaluated in the region of interest using a dense rain gauge network. Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis 3B42, version 6 (TMPA-3B42V6), and 3B42 in real time (TMPA-3B42RT); Global Satellite Mapping of Precipitation (GSMaP) in near–real time (GSMaP-NRT) and Moving Vector with Kalman version (GSMaP-MVK); African Rainfall Estimation Algorithm, version 2.0 (RFE 2.0); Climate Prediction Center (CPC) morphing technique (CMORPH); and Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) are tested and finally corrected using a probability matching method. The corrected products are then used as forcing to the coupled model over the 2003–10 period. The predicted number and size of ponds and their dynamics are greatly improved compared to the model forced only by a single gauge. A more realistic spatiotemporal distribution of the host bite rate of the RVF vectors is thus expected.

This is Lamont-Doherty Earth Observatory contribution 7784.

Corresponding author address: Clement Guilloteau, Observatoire Midi-Pyrénées, Géosciences Environnement Toulouse, 14 avenue Édouard Belin, 31400 Toulouse, France. E-mail: clement.guilloteau@legos.obs-mip.fr

Abstract

Spatiotemporal rainfall variability is a key parameter controlling the dynamics of mosquitoes/vector-borne diseases such as malaria, Rift Valley fever (RVF), or dengue. Impacts from rainfall heterogeneity at small scales (i.e., 1–10 km) on the risk of epidemics (i.e., host bite rate or number of bites per host and per night) must be thoroughly evaluated. A model with hydrological and entomological components for risk prediction of the RVF zoonosis is proposed. The model predicts the production of two mosquito species within a 45 km × 45 km area in the Ferlo region, Senegal. The three necessary steps include 1) best rainfall estimation on a small scale, 2) adequate forcing of a simple hydrological model leading to pond dynamics (ponds are the primary larvae breeding grounds), and 3) best estimate of mosquito life cycles obtained from the coupled entomological model. The sensitivity of the model to the spatiotemporal heterogeneity of rainfall is first tested using high-resolution rain fields from a weather radar. The need for high-resolution rain data is thus demonstrated. Several high-resolution satellite rainfall products are evaluated in the region of interest using a dense rain gauge network. Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis 3B42, version 6 (TMPA-3B42V6), and 3B42 in real time (TMPA-3B42RT); Global Satellite Mapping of Precipitation (GSMaP) in near–real time (GSMaP-NRT) and Moving Vector with Kalman version (GSMaP-MVK); African Rainfall Estimation Algorithm, version 2.0 (RFE 2.0); Climate Prediction Center (CPC) morphing technique (CMORPH); and Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN) are tested and finally corrected using a probability matching method. The corrected products are then used as forcing to the coupled model over the 2003–10 period. The predicted number and size of ponds and their dynamics are greatly improved compared to the model forced only by a single gauge. A more realistic spatiotemporal distribution of the host bite rate of the RVF vectors is thus expected.

This is Lamont-Doherty Earth Observatory contribution 7784.

Corresponding author address: Clement Guilloteau, Observatoire Midi-Pyrénées, Géosciences Environnement Toulouse, 14 avenue Édouard Belin, 31400 Toulouse, France. E-mail: clement.guilloteau@legos.obs-mip.fr

1. Introduction

Various resources in the Sahel (West Africa) depend heavily on rainfall variability: water availability, agricultural yields, and drought/flood risks, all of which have multiple impacts, including public health. The prediction of climate-related health risks is very complex since it involves a long causality chain, with uncertainties for all links. Poor spatiotemporal coverage and poor quality of in situ data are also problems. Rainfall variability has considerable effects on public health through vector-borne diseases such as malaria (the PaluClim project in Burkina Faso), Rift Valley fever (RVF; Rift Valley fever project in Senegal) zoonosis, and dengue fever. In the Ferlo region, Senegal, where the AdaptFVR project took place, RVF is transmitted by two mosquito/vector species: Aedes vexans and Culex poicilipes (Ba et al. 2005). For both species, the principal larvae breeding grounds are intermittent ponds that appear during the Sahelian rainy season from July to October. Ponds are filled and flooded by discrete and convective rainfall events (Lacaux et al. 2007; Gardelle et al. 2010). The clustered pond dynamics depend upon heterogeneous rainfall, and the drainage basin associated with specific pond is small (<1 km2). The water bodies are not very deep (about 1 m), and the total water area during the rainy season has a large interannual variability (Gardelle et al. 2010; Soti et al. 2010).

A 45 km × 45 km area around Barkedji village (15.28°N, 14.87°W) has been delimited for studying and modeling pond–vector interactions. As vector densities are directly linked to pond water surfaces, ponds have been modeled to enable computation of their flooded areas. Vector production is subsequently estimated from the daily variations of these water surfaces. In the studied region, the only in situ measurement of rainfall is provided by a single rain gauge of the Senegalese weather service. In a zone where rainfall is mostly provided by convective systems (local convection and mobile mesoscale convective systems), the representativeness of a single rain gauge over a 2000 km2 area is questionable (Lebel and Amani 1999). High-resolution satellite rainfall data become a valuable input, particularly in regions where the operational rain gauge network is degrading rapidly (Nicholson et al. 2003).

The goal of this paper is to investigate the use of high-resolution satellite rainfall products and see if the prediction of pond dynamics and the RVF vector risk assessment can be improved. Section 2 presents the dataset used for this study. It includes satellite rainfall products and a dense rain gauge network in Niger. In section 3, a newly developed model is presented. Section 4 quantifies the sensitivity of the model to rainfall variability within 45 km × 45 km. It investigates high-resolution satellite rainfall as a forcing field and compares it with using a single rain gauge. Finally, the model-estimated water bodies are compared to series of water surfaces obtained from 15 specific images acquired from the Satellite Pour l’Observation de la Terre 5 (SPOT-5) imagery.

The simplified pond model allows the estimation of water-body areas, while the coupled entomological model estimates emergence, production, and host bite rate (number of bites per host and per night) of the two species of RVF vectors.

2. Data

The datasets that have been used to constrain the model and to test its performance are briefly presented.

a. SPOT-5 images

Fifteen usable (i.e., no cloud coverage) SPOT-5 images (four spectral bands with 10-m resolution) have been taken from 2003 to 2010 over the studied area. A pixel classification enables extraction of water bodies from the images, providing the geographical distribution of ponds (Lacaux et al. 2007). The images also provide a reference dataset for hydrological model output validation since the surface of each individual pond is measured for each given date.

b. Barkedji rain gauge

A rain gauge located in Barkedji (at the center of the study area) since 1964 provides the only long-term daily rainfall time series available in the region. The data from the rain gauge were used as rainfall input for the initial version of the model. In 2010, a second gauge was installed in Niakha, 4 km to the west of the first gauge (see Fig. 1a).

Fig. 1.
Fig. 1.

Hazards mapping on 8 Sep 2003 after the rainy event of 4 Sep 2003. (a) Estimated host bite rate using a uniform rainfall of 29.7 mm for the full area. (b) Estimated host bite rate using GSMaP-MVK spatially distributed rain with same mean value (29.7 mm). The black curves are daily rainfall isohyets. The host bite rate is given in bites per host per night. For this case, mean host bite rates are 8.3 for homogeneous rainfall and 17.5 for heterogeneous rainfall.

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

c. Satellite rainfall estimates

Several satellite products are considered. These satellite products combine infrared observations of cloud-top layers from geostationary satellites and microwave measurement from low-Earth-orbiting (LEO) satellites.

  • Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) rainfall: TMPA-3B42RT is a real-time (RT) rainfall product estimated from a combination of several microwave imagers and geostationary information. TMPA-3B42V6 is a postprocessed version (version 6) of this product that includes a monthly adjustment with rain gauges (Huffman et al. 2007; ftp://disc2.nascom.nasa.gov/data/TRMM/).

  • Global Satellite Mapping of Precipitation (GSMaP) in near–real time (GSMaP-NRT) and Moving Vector with Kalman version (GSMaP-MVK): GSMaP is a Japan Aerospace Exploration Agency (JAXA)–Core Research for Evolutional Science and Technology (CREST) product using a large network of satellites. A near-real-time (NRT) and a postprocessed (MVK) version of the product are available (Aonashi et al. 2009; http://sharaku.eorc.jaxa.jp/GSMaP_crest/).

  • African Rainfall Estimation Algorithm, version 2.0 (RFE 2.0): RFE 2.0, produced by the National Oceanic and Atmospheric Administration’s (NOAA) Climate Prediction Center (CPC), is an Africa-specific real-time product combining satellite and gauge data (Herman et al. 1997; www.cpc.ncep.noaa.gov/products/fews/data.html#rfe2).

  • CPC morphing technique (CMORPH): CMORPH is a real-time product from the Defense Meteorological Satellite Program (DMSP), NOAA, Aqua, and TRMM satellites (Joyce et al. 2004; www.cpc.ncep.noaa.gov/products/janowiak/cmorph_description.html).

  • Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks (PERSIANN): PERSIANN is a product from the Center for Hydrometeorology and Remote Sensing (CHRS), University of California, Irvine, using TRMM and NOAA satellite data (Sorooshian et al. 2000; http://chrs.web.uci.edu/persiann/).

It should be noted that RFE 2.0 has a daily temporal resolution, GSMaP products have an hourly resolution, and all others products have a 3-hourly resolution. In this work, all products are aggregated at the daily resolution, which is the temporal resolution of the model. Spatial resolutions are indicated in Table 1.

Table 1.

Satellite rain products’ daily rainfall rates compared with dense gauges network for the 2003–10 period. The considered criteria are time corr (correlation between the two temporal series), daily rain rates over the network area (1.0° × 1.0°); Rbias, ratio of the 8-yr mean rainfall rates, Rsat/Rground; Res, product nominal spatial resolution; and S1, spatial accordance between satellite and ground estimation. This criterion defines the ability of satellite products to ascertain whether local rain rate is below or above the (spatially) median rain rate. For each rain event (i.e., day for which both satellite and ground rainfall rates are above 2 mm), binary fields are extracted from rain fields by thresholding in regards to daily spatial median rate. Local binary value is 1 if measured rain is above median measured rain and 0 otherwise. S1 ranged from 0 to 1 represents the relative part of agreements between satellite and ground binary rain fields. S2 is the percentage of rain events (rainy days) for which spatial correlation between ground and satellite rain fields (at product’s nominal spatial resolution) is significant. Correlation is considered significant when above 0.3.

Table 1.

d. Dense rain gauge network data from Niamey

Before using satellite products as input to the model, an assessment on their reliability is necessary. Since a single gauge is not a sufficient reference for a comparison, the data from a 55-rain-gauge network covering a 1° × 1° area were used. This reference network located near Niamey (Niger) within the Sahelian climatic regime has been used to evaluate the satellite products and to quantify the biases for potential corrections. The Niamey network (Vischel et al. 2011), which has been operational since 1990, became an important component of the African Monsoon Multidisciplinary Analysis (AMMA) program and is member of the AMMA–Couplage de l’Atmosphère Tropicale Cycle Hydrologique (CATCH) observatory. In the Sahel, it is the only rain gauge network available to provide surface rainfall data down to the convective scale with sufficient accuracy. Mean area values of the daily rainfall over the network, at the resolution of each satellite product, are obtained by block kriging of rain gauge values.

The satellite products were compared to the AMMA-CATCH measured rainfall for the 2003–10 period. In Gosset et al. (2013), the products were compared to the same in situ data for a 1° grid scale. It was concluded that the real-time products display strong positive biases for the Sahelian climate but were able to reproduce the time series of daily rain dynamic. Here, the products are additionally compared to ground measurements at their nominal spatial resolution in order to analyze the spatial information.

The first two columns in Table 1 show that for all the products except GSMaP-NRT, the daily rainfall time series are consistent with those measured by rain gauges over the Niamey region. The temporal correlations (at 1°, daily resolution) are larger than 0.7. This can be compared with the correlation coefficient of one gauge against the spatial mean value obtained by kriging for all other gauges: 0.73 ± 0.04. GSMaP-MVK shows the best results, whereas the performance from GSMaP-NRT is quite poor. The three other real-time products have similar performances for the considered area. Some products (especially PERSIANN) show strong quantitative biases [as also found by Gosset et al. (2013)], but this can be empirically corrected. It is done for all products using a probability matching method (Atlas et al. 1990):
e1
where Rcor is the corrected rain rate, Rsat is the satellite-estimated rain rate, and Fc is a 6° polynomial correction function obtained by fitting the quantile–quantile plot of Rsat distribution against Rground (ground-measured rain rates) distribution. For this correction Rcor, Rsat and Rground are considered at the product’s nominal spatial resolution (Table 1, column 3) with a daily scale. It should be noted that the TMPA-3B42RT and TMPA-3B42V6 products are very similar, 3B42V6 being essentially a bias-corrected version of 3B42RT.

e. Radar rainfall fields from Ouémé (used as a proxy for simulations)

Radar rainfall fields from the Xport radar located in Ouémé (Benin) in 2006 for the AMMA campaign (Gosset et al. 2010; Koffi et al. 2014) are used to constrain the model with high-resolution (i.e., 1 km, every 5 min) rain fields. Vischel et al. (2011) have shown that the spatial structure of the daily rainfall fields in northern Benin and the Sahel in general are very similar. Therefore, the radar estimates are considered as a realistic representation of the small-scale structure of rainfall events found in the Ferlo. The radar rainfall fields aggregated to daily time resolution were used as a proxy to assess the model’s sensitivity to rainfall spatial heterogeneity and resulting sampling errors (see section 4a).

3. The coupled model

The developed tool is at first a simplified hydrological model of ponds within the 45 km × 45 km studied region in the Ferlo region, Senegal. It is then coupled with an entomological model, estimating vectors’ density and host bite rate from the overall water surfaces (Fig. 1a). Previous studies (Ba et al. 2005; Ndiaye et al. 2006) have shown that Aedes and Culex species have a mean flying range of no more than 500 m from the larvae sites during their whole life cycles and that a mosquito’s density decreases linearly with the distance from the ponds. The emergence of the Aedes vexans species is due to a sudden increase in a pond flooded surface. Culex poicilipes species appear later during the season when the pond’s limnimetry is high and stable. Culex density depends on the seasonal rainfall total; each Aedes outburst is related to a specific rainfall event. Aedes outbursts are short in time (about 15 days). There may be several Aedes outbursts per year. On the contrary, Culex temporal distribution is much more continuous, with a unique annual outburst lasting for about 3 months.

a. The hydrological model

The first necessary step to predict vector density is the modeling of ponds’ surface variation. As explained in Lacaux et al. (2007), 1354 ponds have been extracted over the studied domain, from pixel classification using a SPOT-5 image (see section 2a) taken on 26 August 2003. Since the image was taken a few days after several intense rain events, the water surface measured on this image is considered as the theoretical maximum surface Smax for each pond. A 74-mm rain accumulation was measured by the rain gauge in Barkedji (see section 2b) during the 9 previous days and TMPA-3B42V6 gives a 105-mm accumulation for the same period. Given that the average annual accumulation in the region is about 400 mm, 100 mm in 10 days is indeed a very wet period. Onsite observations from local experts (and from the Centre de Suivi Ecologique in Dakar) corroborate the hypothesis that at this given date ponds were at their maximum level. Moreover, none of the subsequent satellite images of the area displayed ponds with larger surfaces.

In a previous version of the model (Vignolles et al. 2010), the daily relative pond surface variation was directly related to daily rainfall amount through an empirical logarithmic function. Here, the water volume provided to each pond by the rain is estimated first, and then the surface variation is calculated through a simplified physical model. It is assumed that for each pond the drainage basin surface is proportional to the maximum pond volume Vmax, which is calculated from the Smax measured from the reference image. The differential water volume provided to a pond by a rainfall event is
e2
where γ is the ratio of drainage basin surface over Vmax and Crunoff is the rainfall amount (millimeters) minus what has percolated through the soil:
e3
e4
where C(i) is the rainfall amount for day i and θ is an empirically determined constant.
Each pond has been arbitrarily modeled as a Gaussian depression with a cylindrical symmetry (Fig. 2): the depth of pond H depends on the distance r to its center according to the equation:
e5
where Rmax is the theoretical maximum radius of a disk of surface Smax and Hmax is the depth of the pond at its center. Since it cannot be measured from satellite images, it is assumed that Hmax = βRmax, where β is a constant empirically determined from topological measurements done locally and for a few ponds.
Fig. 2.
Fig. 2.

Model schematic pond vertical cross section. The pond has a cylindrical symmetry; thus, the water surface is a disk of radius R, with R being the distance from the pond center. The depth H of the pond follows a Gaussian function [(5)]. The gray shading illustrates how a water volume increment dV causes a water height variation dH and a water surface disk radius variation dR.

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

The volume equation is then
e6
where R is the disk water surface radius: .
From these assumptions, a differential water surface dS can be calculated from a differential water volume dV:
e7

Replacing dV in (7) using (2) and (6) eliminates β from the equation. Thus, determining the value of β is not necessary for water surface variation calculation. The calculated water surface is upper limited by the theoretical Smax. The surface volume relation is in accordance with those calculated by Bop et al. (2010) for topological measurements, even for ponds where the real topography is very different from the standard model (Fig. 3). The values of θ, σ, and γ were first (first guess) based on topological measurements and then tuned to minimize the difference between the simulated surfaces and the observed ones (using SPOT-5 images). The model runs on a daily increment.

Fig. 3.
Fig. 3.

Comparison of the surface–volume relation for three ponds estimated from topological measurements by Bop et al. (2010) (dotted lines) with the one from the simple Gaussian shape model of equation: (black solid line).

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

The flushing of ponds by evaporation, percolation, and infiltration is not physically modeled, but a daily flushing rate has been estimated from TerraSAR-X (synthetic aperture radar satellite) and SPOT-5 images observations during dry periods. If day i is a dry day,
e8
where S is the pond water surface in square meters. The flushing rate α was initially estimated to 0.967 (Vignolles et al. 2010). Here, a more specific analysis revealed that α depends also on the size S of the considered water surface. In the current setup,
e9

b. The entomological model

The two mosquito species for the RVF have different life cycles, and thus two different entomological models have been developed.

Aedes females lay eggs on humid ground surrounding the ponds, and the eggs hatch after immersion. The aquatic larval stage lasts 4 days, while the mature Aedes live for around 12 days. Thus, every increase of a pond surface (due to a rainy event) involves Aedes proliferation from 4 days after the rainy event to 15 days after the event. Ndiaye et al. (2006) described the variation of host bite rate during this period. In the model, the daily Aedes host bite rate Ag is computed as follows (Guilloteau 2012):
e10
where i is the indices of the day; Agref is the reference host bite rate variation from Ndiaye et al. (2006); and
e11
where Sa is the surface of the pond after the rainy event, Sb is the surface of the pond before the rainy event, and dSref is the reference surface variation corresponding to the conditions when Agref was established. The variable C2 is an empirical coefficient that depends on the size of the pond.

Culex females lay eggs on aquatic vegetation covering some of the ponds during the summer monsoon. These particular ponds have been detected among others during pixel classification of the 26 August 2003 SPOT-5 image. While Aedes are present at the beginning of the rainy season when the level of the ponds is continuously increasing, Culex appears at the heart of the rainy season when pond water level is sufficiently high and stable for the vegetation to develop. In the model, Culex proliferation is the result of a 4-week (from week iw to week iw + 3) rainfall amount higher than a given threshold TCulex. The Culex model uses weekly increments. The Culex proliferation lasts 12 weeks (from week iw + 6 to week iw + 17) and is proportional to the pond mean surface during the period from iw to iw + 3. Because of these differences in the mosquitoes’ life cycle, only the Aedes outburst has a strong sensitivity to the rainfall small-scale variability; in the rest of the analysis, we therefore concentrate only on Aedes host bite rate.

All 1354 ponds could not be specifically modeled because of the lack of complete topological measurements. Therefore, both hydrological and entomological models were kept simple. Figure 3 shows that using a unique surface–volume relation for all ponds is an approximation. Considering the sparsity of data available for calibration and validation and the large uncertainties on rainfall data inputs (see section 4), establishing a more complex model would not be useful. As an example, the integration of soil type variability using SPOT-5 images has been attempted but with no improvement.

4. Results

This section investigates the model sensitivity to rainfall variability. The improvement of the daily water surfaces estimation using high-resolution satellite rainfall products is quantified. The ability of satellite products to spatialize the estimation of the vectorial risks is shown.

a. Model sensitivity to rainfall sampling uncertainties

Spatiotemporal rainfall heterogeneity renders risk prediction difficult, since a single rain gauge is not sufficient. The influence of parameterization uncertainties and input rainfall data uncertainties have been estimated using Monte Carlo simulations. It leads to the conclusion that, among all sources of model inaccuracy, the daily rainfall rate errors are dominant (Guilloteau 2012). In the current setup, rainfall is provided by a single rain gauge located in Barkedji (see Fig. 1a). As discussed by many authors, rain gauge estimation of rainfall is punctually accurate. Nevertheless, relying upon a single gauge to measure area-averaged rainfall leads to substantial spatial representativeness errors (Lebel and Amani 1999). Forcing the model with data from a single rain gauge is limiting for at least two reasons: 1) uncertainty in the estimation of the mean area-averaged daily rainfall and 2) lack of information on the spatial variability of the rainfall within the studied zone. A numerical test bed is used to analyze the response of the model to the accuracy in the rainfall spatial mean and to rainfall spatial variability.

High-resolution rain fields derived from weather radar were used to simulate the impact of rainfall small-scale spatial variability on a single rain gauge estimation and to deduce the subsequent uncertainty. The data were provided by the Xport radar operating in Benin in 2006 (see section 2e). Two hundred model simulations were run, each time using a single local value of the daily radar rain field at a point selected randomly over the area. This is equivalent to moving a “virtual rain gauge” 200 times. These simulations allow us to quantify the uncertainty associated with using a single point to represent a spatially variable rain field. In Fig. 4, the dispersion of the mean areal-estimated Aedes host bite rate for these simulations is displayed.

Fig. 4.
Fig. 4.

Spatial mean distribution of Aedes host bite rate (bites per host and per night) obtained from 200 simulations. Ordinates are bites per host per night. Abscissa is for days. The blue line is for mean simulated values. The red line is for max simulated values. The shaded green is for the 70% confidence level. Each simulation is equivalent to a random position of a virtual rain gauge network (see text).

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

If the daily estimation of the spatial mean hazard over the studied area is provided by a single rain gauge, the risk variance is very high for reasons explained above. From Fig. 4, it can be seen that for each day of the rainy season, the simulated risk varies from near 0 to very high (more than 250 bites per host and per night) depending on the rain gauge location. The calculated mean coefficient of variation is equal to 3.15. In addition to these simulations, a real case illustration is given: during the 2010 rainy season, a second rain gauge was set up in Niakha, 4 km west of the first one (Fig. 1a). These two datasets have been used as input to the model. Even if the measured rainfall amounts are quite similar with a correlation of 0.79 and total amounts of 450 mm (Barkedji) and 405 mm (Niakha), model-estimated hazards are significantly different, with a daily difference totaling 72% of the mean (see Fig. 5). Thus, a single gauge cannot be used for an accurate hazard estimation over the whole area, even for a 45 km × 45 km area.

Fig. 5.
Fig. 5.

Model-simulated daily series of total water surface (dotted blue line) and spatial mean Aedes host bite rate (solid red line). Simulations using (top) Niakha and (bottom) Barkedji rain gauges as input. Abscissa is for days, from 30 Jul to 12 Nov 2010. Vertical reversed blue lines are gauge-recorded daily rainfall amounts (mm).

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

b. Contribution of satellite products to estimate the total water surface

High-resolution satellite products that combine information from LEO satellites and geostationary information have a good spatial coverage and low spatial sampling error (Roca et al. 2010). Satellite estimation of rainfall is less direct than gauge measurement but it provides a better spatial sampling. Rainfall estimates from high-resolution satellites should therefore permit a better estimation of the spatially average rainfall over the domain.

The products presented in section 2c have been used to constrain the model. The daily pond water surfaces and vector host bite rate have been computed from 2003 to 2010 using each corrected product (section 2c). It was not possible to compare computed host bite rate with a reliable reference dataset since no sufficient in situ measurements of vector host bite rate have been made so far. The water surface of each individual pond for 15 dates between 2003 and 2010 could be estimated from SPOT-5 image pixel classification (see section 2a). The series of model-calculated total water surface and number of active (nondry) ponds have been compared to the measured ones (see Table 2). All simulations with satellite rainfall as input show better correlations with the observations than the one obtained from rain gauge data, both in terms of water surface and number of active ponds.

Table 2.

Model-calculated water surface evaluation using several products as rain input: Comparison of simulated flooded surface and number of active ponds series with SPOT-5 images measured ones. Correlation is , where Xsim and Xobs are the model calculated and measured values for SPOT-5 images values, respectively; mean square error (MSE) is ; and Nash skill score is the criterion taking into account reference signal variance and ranged from −∞ to 1: .

Table 2.

The use of high-resolution satellite rainfall products improves the computation of flooded surface and of the number of ponds estimated compared to the use of a single rain gauge. It should be noted that the GSMaP-MVK product is not available for the entire tested period. Since both real-time and archive products have been used to constrain the model, the latter could explain the relatively low score from the GSMaP simulation when compared with other products. TMPA-3B42RT data are not available for the considered period, so only 3B42V6 data have been tested. CMORPH and PERSIANN simulations show remarkably good scores considering they are real-time products, thus validating the feasibility of the histogram matching method used to correct their biases (section 2c).

c. Impacts of high-resolution rainfall satellite products on spatiotemporal mapping of vectors’ risks

In addition to the rainfall spatial mean, the spatial distribution of rainfall is valuable information to assess the spatial variability of the risk. Since the model is nonlinear, the mean value of the risk over the domain is potentially impacted by rainfall heterogeneity. In other words, an accurate estimate of the rainfall spatial mean amount does not warrant a good estimate of the spatial mean risk.

Because of a small drainage basin, each individual pond is sensitive to local rain and not to the rain spatial mean. The spatial distribution of rainfall within the area is needed to estimate the surface area of each pond. The capacity of the products to provide relevant spatial information within a 45 km × 45 km square is analyzed in the last three columns of Table 1. Once again, GSMaP-MVK displays the best performance, as well as the best spatial resolution. RFE presents an equivalent nominal spatial resolution, but the small-scale information seems less accurate than the GSMaP-MVK one. Unfortunately, GSMaP-MVK is not a real-time product and cannot be used for real-time hazard prediction.

It has been shown previously that the total flooded surface and the number of ponds’ variability are better estimated through the use of high-resolution satellite rainfall products. Figure 6 shows also that the simulated behavior of each individual pond is closer to the observations when satellite data are used. The spatial distribution of the risk is highly heterogeneous, not only because of the geographical distribution of the ponds but also because of the rainfall spatial heterogeneity. If the spatial distribution of the risk within the area is needed, information about the rainfall spatial distribution is a prerequisite. Figure 1b shows that the rainfall distribution has a strong influence on the spatial distribution of the risk and also on the overall hazard intensity. A same amount of mean areal rainfall can be associated with different types of risks depending on how it is spatially distributed. Because of the nonlinearity of the model (mainly when runoff is concerned), heterogeneous rainfall with high local maxima are more prone to cause significant hazards.

Fig. 6.
Fig. 6.

Cumulated distribution of correlation coefficients between observed and simulated water-body surfaces and for the different satellite products. The y axis is for the ponds’ cumulated frequency. Red line is simulated from gauge rain, dark blue is for PERSIANN, light blue is for GSMaP, green is for CMORPH, yellow is for TMPA-3B42V6, and purple is for RFE 2.0.

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

Two ponds with identical maximum surface are modeled exactly the same way. Thus, when a single daily rainfall value is used as input, ponds with identical maximum theoretical surface have identical calculated hazards (Fig. 7). With the use of spatial heterogeneous rainfall input, a pond hazard production depends on its geographical location and not only on its maximum surface. Figure 7 shows the temporal evolution and distribution of the Aedes host bite rate associated with 77 ponds whose maximum surface is around 25 000 m2. When a single daily rainfall value from a rain gauge is used as input for the model, all those simulated ponds behave exactly the same and the calculated Aedes host bite rate associated with all ponds are equal. On the contrary, when a heterogeneous rainfall field is used as input (TMPA-3B42V6 rainfall here), computed host bite rates are very scattered: for example, for the first hazard period (days 7–18) the quantile of 75% is 7 times the median value. Distributions of the daily risks are also very asymmetric, with a positive skew.

Fig. 7.
Fig. 7.

Distribution of Aedes host bite rate in the vicinity of ponds with theoretical surfaces ranging from 0.24 to 0.29 ha. Vertical blue lines are for rainfall events. Abscissa is for days (from 25 Jun to 20 Aug 2003). Estimates using (top) a single rain gauge value and (bottom) TMPA-3B42V6 rainfall values. The box plot (light brown) displays the 25th and 75th quartiles. Small circled dots are for values outside the interquartiles. For TMPA-3B42V6, rainfall values and std dev are plotted. Ordinates are for host bite rate and rainfall values (mm).

Citation: Journal of Hydrometeorology 15, 4; 10.1175/JHM-D-13-0134.1

5. Conclusions

High-resolution satellite rainfall products have been tested as a forcing field for climate- or weather-related health risks in West Africa associated with the zoonosis RVF outburst in the Ferlo region of Senegal. Despite the known biases in the products and the challenge imposed by the small spatiotemporal scales involved, the preliminary results presented here are very promising. Compared to gauge-based estimation, satellite measurements are less affected by spatial sampling errors, which are particularly acute in the Sahel where rainfall is of convective origin and highly variable. The surface area of ponds modeled using satellite rainfall products are much closer to observations than when a single rain gauge is used. The time variation of the pond surface area is also better modeled with satellite rainfall as input. High-resolution satellite rainfall products also provide information about the spatial distribution of the rainfall within the studied area. The fact that the rainfall is spatially heterogeneous impacts the mean risk because of the nonlinear relation between rainfall and runoff. The same amount of rain will generally provide a higher risk if heterogeneously distributed because of higher local maxima (twice as much during 8 September 2003, for example). Spatial information provided by satellite products affects the overall hazard estimation. It also enables us to determine a plausible spatial distribution of the risks. This is very useful to local public health authorities, who can concentrate on preventive actions (such as cattle vaccination and/or larvicide treatment) in high-risk areas (Lafaye et al. 2013). The use of any of the considered rainfall products (once unbiased) is preferable to a single rain gauge estimation for water surface estimation even for a 45 km × 45 km area. The entomological model outputs could not be validated in terms of intraseasonal variations and spatial distribution of mosquitoes because of the lack of in situ measurements. However, we expect improved estimation of surface areas to improve the estimation of the host bite rate. The model is being integrated into an early warning system. Local authorities (Centre de Suivi Ecologique and Direction des Services Vétérinaires du Sénégal) in charge of its operation will continue the validation work. A comparison of model outputs with onsite measurements of seasonal statistics is given in Lafaye et al. (2013).

Acknowledgments

The authors thank the Gestion et Impacts du Changement Climatique (GICC) program supported by the Ministère de l’Ecologie, du Développement Durable des Transports et du Logement (www.gip-ecofor.org/gicc/) and all AdaptFVR project participants. Thanks also to the AMMA-CATCH technical team for the release of the rain gauge data from the Niamey site.

REFERENCES

  • Aonashi, K., and Coauthors, 2009: GSMaP passive, microwave precipitation retrieval algorithm: Algorithm description and validation. J. Meteor. Soc. Japan, 87A, 119136, doi:10.2151/jmsj.87A.119.

    • Search Google Scholar
    • Export Citation
  • Atlas, D., Rosenfeld D. , and Wolf D. B. , 1990: Climatologically tuned reflectivity–rain rate relation and links to area–time integrals. J. Appl. Meteor., 29, 11201135, doi:10.1175/1520-0450(1990)029<1120:CTRRRR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ba, Y., Diallo D. , Kebe C. M. F. , Dia I. , and Diallo M. , 2005: Aspects of bio-ecology of two Rift Valley fever virus vectors in Senegal (West Africa): Aedes vexans and Culex poicilipes. J. Med. Entomol., 42, 739750, doi:10.1603/0022-2585(2005)042[0739:AOBOTR]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bop, M., Soussou S. , Kebe C. M. F. , and Ndione J. A. , 2010: Modélisation du fonctionnement hydrologique d’un bassin endoréique pour une application à l’étude de la fièvre de la Vallée du Rift (FVR). IAHS Publ.,340, 305–313. [Available online at http://iahs.info/uploads/dms/abs_340_0305.pdf.]

  • Gardelle, J., Hiernaux P. , Kergoat L. , and Grippa M. , 2010: Less rain, more water in ponds: A remote sensing study of the dynamics of surface waters from 1950 to present in pastoral Sahel (Gourma region, Mali). Hydrol. Earth Syst. Sci., 14, 309324, doi:10.5194/hess-14-309-2010.

    • Search Google Scholar
    • Export Citation
  • Gosset, M., Zahiri E. P. , and Moumouni S. , 2010: Rain drop size distributions variability and impact on X-band polarimetric radar retrieval: Results from the AMMA campaign in Benin. Quart. J. Roy. Meteor. Soc., 136, 243256, doi:10.1002/qj.556.

    • Search Google Scholar
    • Export Citation
  • Gosset, M., Viarre J. , Quantin G. , and Alcoba M. , 2013: Evaluation of several rainfall products used for hydrological applications over West Africa using two high-resolution gauge networks. Quart. J. Roy. Meteor. Soc., 139, 923940, doi:10.1002/qj.2130.

    • Search Google Scholar
    • Export Citation
  • Guilloteau, C., 2012: Apport des Outils Spatiaux au Suivi de la Fièvre de la Vallée du Rift: Estimation et Prévision des Densités de Vecteurs. Master Rep., Laboratoire d’Aérologie, Observatoire Midi-Pyrénée, Toulouse, France, 58 pp. [Available online at http://temis.documentation.developpement-durable.gouv.fr/documents/Temis/0077/Temis-0077972/20657_B.pdf.]

  • Herman, A., Kumar V. B. , Arkin P. A. , and Kousky J. V. , 1997: Objectively determined 10-day African rainfall estimates created for famine early warning. Int. J. Remote Sens., 18, 21472159, doi:10.1080/014311697217800.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., Adler R. F. , Bolvin D. T. , Gu G. , Nelkin E. J. , Bowman K. P. , Hong Y. , Stocker E. P. , and Wolff D. B. , 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 3855, doi:10.1175/JHM560.1.

    • Search Google Scholar
    • Export Citation
  • Joyce, R. J., Janowiak J. E. , Arkin P. A. , and Xie P. , 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487503, doi:10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Koffi, A. K., Gosset M. , Zahiri E.-P. , Ochou A. D. , Kacou M. , Cazenave F. , and Assamoi P. , 2014: Evaluation of X-band polarimetric radar estimation of rainfall and rain drop size distribution parameters in West Africa. Atmos. Res., 143, 438461, doi:10.1016/j.atmosres.2014.03.009.

    • Search Google Scholar
    • Export Citation
  • Lacaux, J.-P., Tourre Y. M. , Vignolles C. , Ndione J.-A. , and Lafaye M. , 2007: Classification of ponds from high-spatial resolution remote sensing: Application to Rift Valley fever epidemics in Senegal. Remote Sens. Environ., 106, 6674, doi:10.1016/j.rse.2006.07.012.

    • Search Google Scholar
    • Export Citation
  • Lafaye, M., and Coauthors, 2013: Rift Valley fever dynamics in Senegal: A project for pro-active adaptation and improvement of livestock raising management. Geospat. Health, 8, 279288.

    • Search Google Scholar
    • Export Citation
  • Lebel, T., and Amani A. , 1999: Rainfall estimation in the Sahel: What is the ground truth? J. Appl. Meteor., 38, 555568, doi:10.1175/1520-0450(1999)038<0555:REITSW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ndiaye, P. I., Bicout D. J. , Mondet B. , and Sabatier P. , 2006: Rainfall triggered dynamics of Aedes mosquito aggressiveness. J. Theor. Biol., 243, 222229, doi:10.1016/j.jtbi.2006.06.005.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., and Coauthors, 2003: Validation of TRMM and other rainfall estimates with a high-density gauge dataset for West Africa. Part II: Validation of TRMM rainfall products. J. Appl. Meteor., 42, 13551368, doi:10.1175/1520-0450(2003)042<1355:VOTAOR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Roca, R., Chambon P. , Jobard I. , Kirstetter P. E. , Gosset M. , and Berges J. C. , 2010: Comparing satellite and surface rainfall products over West Africa at meteorologically relevant scales during the AMMA campaign using error estimates. J. Appl. Meteor. Climatol., 49, 715731, doi:10.1175/2009JAMC2318.1.

    • Search Google Scholar
    • Export Citation
  • Sorooshian, S., Hsu K.-L. , Gao X. , Gupta H. V. , Imam B. , and Braithwaite D. , 2000: Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Amer. Meteor. Soc., 81, 20352046, doi:10.1175/1520-0477(2000)081<2035:EOPSSE>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Soti, V., Puech C. , Lo Seen D. , Bertan A. , Vignolles C. , Mondet B. , Dessay N. , and Tran A. , 2010: The potential for remote sensing and hydrologic modelling to assess the spatio-temporal dynamics of ponds in the Ferlo Region (Senegal). Hydrol. Earth Syst. Sci., 14, 14491464, doi:10.5194/hess-14-1449-2010.

    • Search Google Scholar
    • Export Citation
  • Vignolles, C., Tourre Y. M. , Mora O. , and Imanache L. , 2010: TerraSAR-X high resolution radar remote sensing: An operational warning system for Rift Valley fever risk. Geospat. Health, 5, 2331.

    • Search Google Scholar
    • Export Citation
  • Vischel, T., Quantin G. , Lebel T. , Viarre J. , Gosset M. , Cazenave F. , and Panthou G. , 2011: Generation of high-resolution rain fields in West Africa: Evaluation of dynamic interpolation methods. J. Hydrometeor., 12, 1465–1482, doi:10.1175/JHM-D-10-05015.1.

    • Search Google Scholar
    • Export Citation
Save
  • Aonashi, K., and Coauthors, 2009: GSMaP passive, microwave precipitation retrieval algorithm: Algorithm description and validation. J. Meteor. Soc. Japan, 87A, 119136, doi:10.2151/jmsj.87A.119.

    • Search Google Scholar
    • Export Citation
  • Atlas, D., Rosenfeld D. , and Wolf D. B. , 1990: Climatologically tuned reflectivity–rain rate relation and links to area–time integrals. J. Appl. Meteor., 29, 11201135, doi:10.1175/1520-0450(1990)029<1120:CTRRRR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ba, Y., Diallo D. , Kebe C. M. F. , Dia I. , and Diallo M. , 2005: Aspects of bio-ecology of two Rift Valley fever virus vectors in Senegal (West Africa): Aedes vexans and Culex poicilipes. J. Med. Entomol., 42, 739750, doi:10.1603/0022-2585(2005)042[0739:AOBOTR]2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bop, M., Soussou S. , Kebe C. M. F. , and Ndione J. A. , 2010: Modélisation du fonctionnement hydrologique d’un bassin endoréique pour une application à l’étude de la fièvre de la Vallée du Rift (FVR). IAHS Publ.,340, 305–313. [Available online at http://iahs.info/uploads/dms/abs_340_0305.pdf.]

  • Gardelle, J., Hiernaux P. , Kergoat L. , and Grippa M. , 2010: Less rain, more water in ponds: A remote sensing study of the dynamics of surface waters from 1950 to present in pastoral Sahel (Gourma region, Mali). Hydrol. Earth Syst. Sci., 14, 309324, doi:10.5194/hess-14-309-2010.

    • Search Google Scholar
    • Export Citation
  • Gosset, M., Zahiri E. P. , and Moumouni S. , 2010: Rain drop size distributions variability and impact on X-band polarimetric radar retrieval: Results from the AMMA campaign in Benin. Quart. J. Roy. Meteor. Soc., 136, 243256, doi:10.1002/qj.556.

    • Search Google Scholar
    • Export Citation
  • Gosset, M., Viarre J. , Quantin G. , and Alcoba M. , 2013: Evaluation of several rainfall products used for hydrological applications over West Africa using two high-resolution gauge networks. Quart. J. Roy. Meteor. Soc., 139, 923940, doi:10.1002/qj.2130.

    • Search Google Scholar
    • Export Citation
  • Guilloteau, C., 2012: Apport des Outils Spatiaux au Suivi de la Fièvre de la Vallée du Rift: Estimation et Prévision des Densités de Vecteurs. Master Rep., Laboratoire d’Aérologie, Observatoire Midi-Pyrénée, Toulouse, France, 58 pp. [Available online at http://temis.documentation.developpement-durable.gouv.fr/documents/Temis/0077/Temis-0077972/20657_B.pdf.]

  • Herman, A., Kumar V. B. , Arkin P. A. , and Kousky J. V. , 1997: Objectively determined 10-day African rainfall estimates created for famine early warning. Int. J. Remote Sens., 18, 21472159, doi:10.1080/014311697217800.

    • Search Google Scholar
    • Export Citation
  • Huffman, G. J., Adler R. F. , Bolvin D. T. , Gu G. , Nelkin E. J. , Bowman K. P. , Hong Y. , Stocker E. P. , and Wolff D. B. , 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 3855, doi:10.1175/JHM560.1.

    • Search Google Scholar
    • Export Citation
  • Joyce, R. J., Janowiak J. E. , Arkin P. A. , and Xie P. , 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487503, doi:10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Koffi, A. K., Gosset M. , Zahiri E.-P. , Ochou A. D. , Kacou M. , Cazenave F. , and Assamoi P. , 2014: Evaluation of X-band polarimetric radar estimation of rainfall and rain drop size distribution parameters in West Africa. Atmos. Res., 143, 438461, doi:10.1016/j.atmosres.2014.03.009.

    • Search Google Scholar
    • Export Citation
  • Lacaux, J.-P., Tourre Y. M. , Vignolles C. , Ndione J.-A. , and Lafaye M. , 2007: Classification of ponds from high-spatial resolution remote sensing: Application to Rift Valley fever epidemics in Senegal. Remote Sens. Environ., 106, 6674, doi:10.1016/j.rse.2006.07.012.

    • Search Google Scholar
    • Export Citation
  • Lafaye, M., and Coauthors, 2013: Rift Valley fever dynamics in Senegal: A project for pro-active adaptation and improvement of livestock raising management. Geospat. Health, 8, 279288.

    • Search Google Scholar
    • Export Citation
  • Lebel, T., and Amani A. , 1999: Rainfall estimation in the Sahel: What is the ground truth? J. Appl. Meteor., 38, 555568, doi:10.1175/1520-0450(1999)038<0555:REITSW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ndiaye, P. I., Bicout D. J. , Mondet B. , and Sabatier P. , 2006: Rainfall triggered dynamics of Aedes mosquito aggressiveness. J. Theor. Biol., 243, 222229, doi:10.1016/j.jtbi.2006.06.005.

    • Search Google Scholar
    • Export Citation
  • Nicholson, S. E., and Coauthors, 2003: Validation of TRMM and other rainfall estimates with a high-density gauge dataset for West Africa. Part II: Validation of TRMM rainfall products. J. Appl. Meteor., 42, 13551368, doi:10.1175/1520-0450(2003)042<1355:VOTAOR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Roca, R., Chambon P. , Jobard I. , Kirstetter P. E. , Gosset M. , and Berges J. C. , 2010: Comparing satellite and surface rainfall products over West Africa at meteorologically relevant scales during the AMMA campaign using error estimates. J. Appl. Meteor. Climatol., 49, 715731, doi:10.1175/2009JAMC2318.1.

    • Search Google Scholar
    • Export Citation
  • Sorooshian, S., Hsu K.-L. , Gao X. , Gupta H. V. , Imam B. , and Braithwaite D. , 2000: Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Amer. Meteor. Soc., 81, 20352046, doi:10.1175/1520-0477(2000)081<2035:EOPSSE>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Soti, V., Puech C. , Lo Seen D. , Bertan A. , Vignolles C. , Mondet B. , Dessay N. , and Tran A. , 2010: The potential for remote sensing and hydrologic modelling to assess the spatio-temporal dynamics of ponds in the Ferlo Region (Senegal). Hydrol. Earth Syst. Sci., 14, 14491464, doi:10.5194/hess-14-1449-2010.

    • Search Google Scholar
    • Export Citation
  • Vignolles, C., Tourre Y. M. , Mora O. , and Imanache L. , 2010: TerraSAR-X high resolution radar remote sensing: An operational warning system for Rift Valley fever risk. Geospat. Health, 5, 2331.

    • Search Google Scholar
    • Export Citation
  • Vischel, T., Quantin G. , Lebel T. , Viarre J. , Gosset M. , Cazenave F. , and Panthou G. , 2011: Generation of high-resolution rain fields in West Africa: Evaluation of dynamic interpolation methods. J. Hydrometeor., 12, 1465–1482, doi:10.1175/JHM-D-10-05015.1.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Hazards mapping on 8 Sep 2003 after the rainy event of 4 Sep 2003. (a) Estimated host bite rate using a uniform rainfall of 29.7 mm for the full area. (b) Estimated host bite rate using GSMaP-MVK spatially distributed rain with same mean value (29.7 mm). The black curves are daily rainfall isohyets. The host bite rate is given in bites per host per night. For this case, mean host bite rates are 8.3 for homogeneous rainfall and 17.5 for heterogeneous rainfall.

  • Fig. 2.

    Model schematic pond vertical cross section. The pond has a cylindrical symmetry; thus, the water surface is a disk of radius R, with R being the distance from the pond center. The depth H of the pond follows a Gaussian function [(5)]. The gray shading illustrates how a water volume increment dV causes a water height variation dH and a water surface disk radius variation dR.

  • Fig. 3.

    Comparison of the surface–volume relation for three ponds estimated from topological measurements by Bop et al. (2010) (dotted lines) with the one from the simple Gaussian shape model of equation: (black solid line).

  • Fig. 4.

    Spatial mean distribution of Aedes host bite rate (bites per host and per night) obtained from 200 simulations. Ordinates are bites per host per night. Abscissa is for days. The blue line is for mean simulated values. The red line is for max simulated values. The shaded green is for the 70% confidence level. Each simulation is equivalent to a random position of a virtual rain gauge network (see text).

  • Fig. 5.

    Model-simulated daily series of total water surface (dotted blue line) and spatial mean Aedes host bite rate (solid red line). Simulations using (top) Niakha and (bottom) Barkedji rain gauges as input. Abscissa is for days, from 30 Jul to 12 Nov 2010. Vertical reversed blue lines are gauge-recorded daily rainfall amounts (mm).

  • Fig. 6.

    Cumulated distribution of correlation coefficients between observed and simulated water-body surfaces and for the different satellite products. The y axis is for the ponds’ cumulated frequency. Red line is simulated from gauge rain, dark blue is for PERSIANN, light blue is for GSMaP, green is for CMORPH, yellow is for TMPA-3B42V6, and purple is for RFE 2.0.

  • Fig. 7.

    Distribution of Aedes host bite rate in the vicinity of ponds with theoretical surfaces ranging from 0.24 to 0.29 ha. Vertical blue lines are for rainfall events. Abscissa is for days (from 25 Jun to 20 Aug 2003). Estimates using (top) a single rain gauge value and (bottom) TMPA-3B42V6 rainfall values. The box plot (light brown) displays the 25th and 75th quartiles. Small circled dots are for values outside the interquartiles. For TMPA-3B42V6, rainfall values and std dev are plotted. Ordinates are for host bite rate and rainfall values (mm).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1348 881 56
PDF Downloads 185 53 0