Anton Vrieling

Length of Growing Period over Africa: Variability and Trends from 30 Years of NDVI Time Series

The spatial distribution of crops and farming systems in Africa is determined by the duration of the period during which crop and livestock water requirements are met. The length of growing period (LGP) is normally assessed from weather station data—scarce in large parts of Africa—or coarse-resolution rainfall estimates derived from weather satellites. In this study, we analyzed LGP and its variability based on the 1981–2011 GIMMS NDVI3g dataset. We applied a variable threshold method in combination with a searching algorithm to determine start- and end-of-season. We obtained reliable LGP estimates for arid, semi-arid and sub-humid climates that are consistent in space and time. This approach effectively mapped bimodality for clearly separated wet seasons in the Horn of Africa. Due to cloud contamination, the identified bimodality along the Guinea coast was judged to be less certain. High LGP variability is dominant in arid and semi-arid areas, and is indicative of crop failure risk. Significant negative trends in LGP were found for the northern part of the Sahel, for parts of Tanzania and northern Mozambique, and for the short rains of eastern Kenya. Positive trends occurred across western Africa, in southern Africa, and in eastern Kenya for the long rains. Our LGP analysis provides useful information for the mapping of farming systems, and to study the effects of climate variability and other drivers of change on vegetation and crop suitability.

Timing of Erosion and Satellite Data: A Multi-resolution Approach to Soil Erosion Risk Mapping

Abstract Erosion reduces soil productivity and causes negative downstream impacts. Erosion processes occur on areas with erodible soils and sloping terrain when high-intensity rainfall coincides with limited vegetation cover. Timing of erosion events has implications on the selection of satellite imagery, used to describe spatial patterns of protective vegetation cover. This study proposes a method for erosion risk mapping with multi-temporal and multi-resolution satellite data. The specific objectives of the study are: (1) to determine when during the year erosion risk is highest using coarse-resolution data, and (2) to assess the optimal timing of available medium-resolution images to spatially represent vegetation cover during the high erosion risk period. Analyses were performed for a 100-km2 pasture area in the Brazilian Cerrados. The first objective was studied by qualitatively comparing three-hourly TRMM rainfall estimates with MODIS NDVI time series for one full year (August 2002–August 2003). November and December were identified as the months with highest erosion risk. The second objective was examined with a time series of six available ASTER images acquired in the same year. Persistent cloud cover limited image acquisition during high erosion risk periods. For each ASTER image the NDVI was calculated and classified into five equally sized classes. Low NDVI was related to high erosion risk and vice versa. A DEM was used to set approximately flat zones to very low erosion risk. The six resulting risk maps were compared with erosion features, visually interpreted from a fine-resolution QuickBird image. Results from the October ASTER image gave highest accuracy (84%), showing that erosion risk mapping in the Brazilian Cerrados can best be performed with images acquired shortly before the first erosion events. The presented approach that uses coarse-resolution temporal data for determining erosion periods and medium-resolution data for effective erosion risk mapping is fast and straightforward. It shows good potential for successful application in other areas with high spatial and temporal variability of vegetation cover.

Migration and environment in Ghana: a cross-district analysis of human mobility and vegetation dynamics

Migration—environment linkages are at the centre of media attention because of public concern about climate change and a perceived “flooding” of migrants from less developed countries into more affluent parts of the world. In the past few years, a substantial body of conceptual literature about environmentally induced migration has evolved, but there is still a paucity of empirical work in this area. Moreover, the environmental causes of migration have been studied largely in isolation of the environmental consequences. In this paper we present an analysis of migration and vegetation dynamics for one country (Ghana) to assess four migration—environment linkages. On the one hand, we look at two environmental drivers of migration: environmental push and pull. On the other hand, we look at the environmental impact of migration on source and destination areas. Census data at the district level (N=110) are used to map domestic migration flows in Ghana, which are then related to vegetation dynamics retrieved from a remotely sensed Normalized Difference Vegetation Index (NDVI) dataset (1981— 2006). The analysis shows that at the national level, there are significant but weak correlations between migration and vegetation cover and trends therein. Districts with a migration deficit (more out-migration than in-migration) tend to be more sparsely vegetated and have experienced a more positive NDVI trend over the past quarter century than districts with a migration surplus. A disaggregation of data in three principle migration systems shows stronger correlations. Namely that north—south migration and cocoa frontier settlement have important environmental dimensions, but environmental factors do not seem to play a major role in migration to the capital, Accra. An important insight from this paper is that migration flows in Ghana can be explained partly by vegetation dynamics but are also strongly related to rural population densities. This is because access to natural resources is often more important than the scarcity or abundance of natural resources per se. This study further shows that satellite remote sensing can provide valuable input to analyses of migration—environment linkages.

Neural networks as a tool for constructing continuous NDVI time series from AVHRR and MODIS

The long term Advanced Very High Resolution Radiometer (AVHRR)‐Normalized Difference Vegetation Index (NDVI) record provides a critical historical perspective on vegetation dynamics necessary for global change research. Despite the proliferation of new sources of global, moderate resolution vegetation datasets, the remote sensing community is still struggling to create datasets derived from multiple sensors that allow the simultaneous use of spectral vegetation for time series analysis. To overcome the non‐stationary aspect of NDVI, we use an artificial neural network (ANN) to map the NDVI indices from AVHRR to those from MODIS using atmospheric, surface type and sensor‐specific inputs to account for the differences between the sensors. The NDVI dynamics and range of MODIS NDVI data at 1° is matched and extended through the AVHRR record. Four years of overlap between the two sensors is used to train a neural network to remove atmospheric and sensor specific effects on the AVHRR NDVI. In this paper, we present the resulting continuous dataset, its relationship to MODIS data, and a validation of the product.

Recent trends in agricultural production of Africa based on AVHRR NDVI time series

African agriculture is expected to be hard-hit by ongoing climate change. Effects are heterogeneous within the continent, but in some regions resulting production declines have already impacted food security. Time series of remote sensing data allow us to examine where persistent changes occur. In this study, we propose to examine recent trends in agricultural production using 26 years of NDVI data. We use the 8-km resolution AVHRR NDVI 15-day composites of the GIMMS group (1981-2006). Temporal data-filtering is applied using an iterative Savitzky-Golay algorithm to remove noise in the time series. Except for some regions with persistent cloud cover, this filter produced smooth profiles. Subsequently two methods were used to extract phenology indicators from the profiles for each raster cell. These indicators include start of season, length of season, time of maximum NDVI, maximum NDVI, and cumulated NDVI over the season. Having extracted the indicators for every year, we aggregate them for agricultural areas at sub-national level using a crop mask. The aggregation was done to focus the analysis on agriculture, and allow future comparison with yield statistics. Trend analysis was performed for yearly aggregated indicators to assess where persistent change occurred during the 26-year period. Results show that the phenology extraction method chosen has an important influence on trend outcomes. Consistent trends suggest a rising yield trend for 500-1100 mm rainfall zones ranging from Senegal to Sudan. Negative yield trends are expected for the southern Atlantic coast of West Africa, and for western Tanzania.

Predictive Factors and Risk Mapping for Rift Valley Fever Epidemics in Kenya.

Background To-date, Rift Valley fever (RVF) outbreaks have occurred in 38 of the 69 administrative districts in Kenya. Using surveillance records collected between 1951 and 2007, we determined the risk of exposure and outcome of an RVF outbreak, examined the ecological and climatic factors associated with the outbreaks, and used these data to develop an RVF risk map for Kenya. Methods Exposure to RVF was evaluated as the proportion of the total outbreak years that each district was involved in prior epizootics, whereas risk of outcome was assessed as severity of observed disease in humans and animals for each district. A probability-impact weighted score (1 to 9) of the combined exposure and outcome risks was used to classify a district as high (score ≥ 5) or medium (score ≥2 - <5) risk, a classification that was subsequently subjected to expert group analysis for final risk level determination at the division levels (total = 391 divisions). Divisions that never reported RVF disease (score < 2) were classified as low risk. Using data from the 2006/07 RVF outbreak, the predictive risk factors for an RVF outbreak were identified. The predictive probabilities from the model were further used to develop an RVF risk map for Kenya. Results The final output was a RVF risk map that classified 101 of 391 divisions (26%) located in 21 districts as high risk, and 100 of 391 divisions (26%) located in 35 districts as medium risk and 190 divisions (48%) as low risk, including all 97 divisions in Nyanza and Western provinces. The risk of RVF was positively associated with Normalized Difference Vegetation Index (NDVI), low altitude below 1000m and high precipitation in areas with solonertz, luvisols and vertisols soil types (p <0.05). Conclusion RVF risk map serves as an important tool for developing and deploying prevention and control measures against the disease.

An auto-calibration procedure for empirical solar radiation models

Solar radiation data are an important input for estimating evapotranspiration and modelling crop growth. Direct measurement of solar radiation is now carried out in most European countries, but the network of measuring stations is too sparse for reliable interpolation of measured values. Instead of direct measurements, solar radiation may be estimated from empirical solar radiation models that employ more commonly measured variables or direct outputs of general and regional circulation models (such as air temperature). Coefficients for these models are site-dependent. This usually implies that they are estimated for stations with direct radiation measurements, but need to be interpolated for other locations. In this paper, we introduce a procedure to auto-calibrate empirical solar radiation models that are based on daily air temperature range, i.e. Bristow and Campbell (1984), and Hargreaves et?al. (1985). Meteosat Second Generation data were used to create two static look-up tables of mean cloud cover and clear-sky transmissivity as input for the auto-calibration procedure. We demonstrate that daily solar radiation can be accurately estimated from daily air temperature range measurements without site-specific empirical coefficients that require stations that measure solar radiation. The average relative root mean square error for our auto-calibrated models was comparable to ground-measurement-based calibration; only 1% higher for the Bristow and Campbell model (p?<?0.05, n?=?126), and 2% higher for the Hargreaves model (p?<?0.05, n?=?126). The mean bias error, relative mean bias error and the slope of linear regression were not statistically different in comparison to ground-measurement-based calibration for the Bristow and Campbell model. When our new solar radiation retrieval algorithm is used to estimate evapotranspiration, we found similar accuracies when using solar radiation input from ground- and auto-calibration. We conclude that our auto-calibration procedure results in accurate solar radiation retrievals, and requires only daily air temperature time series as input. The same procedure could easily be applied to other empirical solar radiation models. We calibrated two solar radiation models for European stations from air temperature data.No solar radiation measurements were needed for the auto-calibration procedure.Similar accuracies as compared to calibration with ground measurements were achieved.The auto-calibration procedure is applicable to other solar radiation models.A freely-available software implementation of the calibration procedure is provided.

Spatially detailed retrievals of spring phenology from single-season high-resolution image time series

Vegetation indices derived from satellite image time series have been extensively used to estimate the timing of phenological events like season onset. Medium spatial resolution (≥250 m) satellite sensors with daily revisit capability are typically employed for this purpose. In recent years, phenology is being retrieved at higher resolution (≤30 m) in response to increasing availability of high-resolution satellite data. To overcome the reduced acquisition frequency of such data, previous attempts involved fusion between high- and medium-resolution data, or combinations of multi-year acquisitions in a single phenological reconstruction. The objectives of this study are to demonstrate that phenological parameters can now be retrieved from single-season high-resolution time series, and to compare these retrievals against those derived from multi-year high-resolution and single-season medium-resolution satellite data. The study focuses on the island of Schiermonnikoog, the Netherlands, which comprises a highly-dynamic saltmarsh, dune vegetation, and agricultural land. Combining NDVI series derived from atmospherically-corrected images from RapidEye (5 m-resolution) and the SPOT5 Take5 experiment (10m-resolution) acquired between March and August 2015, phenological parameters were estimated using a function fitting approach. We then compared results with phenology retrieved from four years of 30 m Landsat 8 OLI data, and single-year 100 m Proba-V and 250 m MODIS temporal composites of the same period. Retrieved phenological parameters from combined RapidEye/SPOT5 displayed spatially consistent results and a large spatial variability, providing complementary information to existing vegetation community maps. Retrievals that combined four years of Landsat observations into a single synthetic year were affected by the inclusion of years with warmer spring temperatures, whereas adjustment of the average phenology to 2015 observations was only feasible for a few pixels due to cloud cover around phenological transition dates. The Proba-V and MODIS phenology retrievals scaled poorly relative to their high-resolution equivalents, indicating that medium-resolution phenology retrievals need to be interpreted with care, particularly in landscapes with fine-scale land cover variability.

Evaluation of the Standardized Precipitation Index as an early predictor of seasonal vegetation production anomalies in the Sahel

ABSTRACT We analysed the performance and timeliness of the Standardized Precipitation Index (SPI) in anticipating deviations from mean seasonal vegetation productivity in the Sahel. Gridded rainfall estimates are used to compute the SPI for 1–6-month timescales, whereas the Z-score of the cumulative value of the Fraction of Absorbed Photosynthetically Active Radiation over the growing season (zCFAPAR) is used as a proxy of seasonal productivity. Results show that the strength of the link varies in space as a function of both the SPI timescale and the timing of the SPI calculation with respect to the vegetative season’s progress. For productivity forecasting, we propose an operational strategy to select per grid cell the SPI timescale and computation time with the highest correlation with zCFAPAR at different moments of the season. The linear relationship between SPI and zCFAPAR is significant for 32–66% of the study area, depending on the timing at which SPI is considered (at 0% and 75% of the seasonal progress, respectively). For these areas, the selected SPI explains on average about 40% of the variance of zCFAPAR and may thus assist in the earlier identification of agricultural drought.

Constructing boundary-consistent population time series for the municipalities of the Netherlands, 1988–2011

Frequent spatial reorganization of administrative units is common in many countries. It may comprise the merging or division of spatial units, or boundary changes between units. These reorganizations prevent the effective assessment of longer-term population dynamics at a detailed spatial level. To deal with this problem in the Netherlands, we developed a new temporal correction method for the populations of municipalities. Rather than estimating the affected population, we used existing data on the number of persons affected by each spatial change. We assumed that before any boundary changes took place, population development was spatially uniform within a municipality. Systematically transferring proportions of the population from original to newly defined target municipalities back in time provided a corrected time series for 1988–2011, based on the 2011 municipal boundaries. Overall, our results correspond well with a detailed reconstruction for 1999–2009 based on data for individual households. Our procedure may be applicable in other countries with effective population registration systems.