To read the full-text of this research, you can request a copy directly from the authors.
... In addition, most of the maize produced in the developing world comes from low income countries, with the livelihoods of the most vulnerable populations strongly dependent on maize production and its fluctuations (Bassu et al., 2014;Shiferaw et al., 2011). In West Africa, maize as a staple crop plays a central role in fulfilling population food requirements (Chivasa et al., 2017;Shiferaw et al., 2011), which were estimated at more than 30 kg/capita/year in 2013 by FAOSTAT. However, important climate and demographic trends combine to worsen an already difficult situation in There are several ways of estimating crop yields, from a plot to continent scale. ...
... For highland farming systems characterized by rugged terrain, very small fields, a large variety of crop species and/or the presence of mixed crops or agroforestery systems, such results probably could not be achieved without an adaptation of the proposed approach. Again, a lot of hope is placed in the new generation of high-resolution sensors, as well as the improvement of spatial disaggregation techniques for low-resolution product sensors, such as SMOS, making it possible to mitigate the impact of mixed-pixels on the spectral signatures of cropping areas and thus to improve crop yield estimating and forecasting in heterogeneous African farming systems (Chivasa et al., 2017). ...
Remote sensing data, crop modelling, and statistical methods are combined in an original method to overcome current limitations of crop yield estimation. It is then tested for timely estimation of maize grain yields and their year-on-year variability in Burkina Faso. Outputs from the SARRA-O crop model were used as a proxy for observed data for calibration. The final remote sensing-based yield model was constructed on the interaction between aboveground biomass at flowering (AGB-F) and crop water stress (Cstr) over the reproductive and maturation phases. Various vegetation and drought-related indices were derived from different spectral domains and tested. Model performance was evaluated by cross-validation against (a) simulated yields and (b) independent yields from ground surveys aggregated at village level. The results showed that the RF (Random Forest) model outperformed the MLR (Multiple Linear Regression) model for year-on-year yield estimation at the end of the season when compared to simulated yields. Surface soil moisture (SSM) information, as a proxy for soil water available for plant growth, together with information on the temperature of the canopy cover, helped to improve the RF maize yield model, impacting more particularly the estimation of crop water stress. Lastly, two months before harvest the RF model predicted 46% of the observed end-of-season maize grain yield variability. The combined remote sensing, crop model and machine learning method is thus an effective approach for estimating and forecasting inter-annual maize crop yields in environments where field data are scarce, such as in most parts of the African continent. However, more research is needed to better retrieve the spatial variability of yields in order to strengthen current agricultural monitoring systems, and to address societal challenges, such as declining food security.
... EO-derived yield predictors, on the other hand, evaluate actual precursors of yield and can be applied at granular scales increasingly congruent with the reality of smallholder agricultural systems. A set of recent studies (Schut et al., 2018;Jin et al., 2017;Han et al., 2017;Burke and Lobell, 2017;Jain et al., 2016) focused on yield estimation from EO. Chivasa et al. (2017) suggested that the decreasing cost of EO data opens opportunities for maize yield estimation in heterogeneous African agricultural landscapes, but that reliable field data and accurate classification methods remain key. Jin et al. (2017) compared an empirically calibrated model using yield measurements and an uncalibrated model based on pseudo training data generated from a crop model. ...
In Mali's cotton belt, spatial variability in management practices, soil fertility and rainfall strongly impact crop productivity and the livelihoods of smallholder farmers. To identify crop growth conditions and hence improve food security, accurate assessment of local crop production is key. However, production estimates in heterogeneous smallholder farming systems often rely on labor-intensive surveys that are not easily scalable, nor exhaustive. Recent advances in high-resolution earth observation (EO) open up new possibilities to work in heterogeneous smallholder systems. This paper develops a method to estimate individual crop production at farm-to-community scales using high-resolution Sentinel-2 time series and ground data in the commune of Koningue, Mali. Our estimation of agricultural production relies on (i) a supervised, pixel-based crop type classification inside an existing cropland mask, (ii) a comparison of yield estimators based on spectral indices and derived leaf area index (LAI), and (iii) a Monte Carlo approach combining the resulting unbiased crop area estimate and the uncertainty on the associated yield estimate. Results show that crop types can be mapped from Sentinel-2 data with 80% overall accuracy (OA), with best performances observed for cotton (Fscore 94%), maize (88%) and millet (83%), while peanut (71%) and sorghum (46%) achieve less. Incorporation of parcel limits extracted from very high-resolution imagery is shown to increase OA to 85%. Obtained through inverse radiative transfer modeling, Sen2-Agri estimates of LAI achieve better prediction of final grain yield than various vegetation indices, reaching R 2 of 0.68, 0.62, 0.8 and 0.48 for cotton, maize, millet and sorghum respectively. The uncertainty of Monte Carlo production estimates does not exceed 0.3% of the total production for each crop type.
... Recent advances in sensor technology and the availability of free high-resolution (spatial and temporal) multispectral satellite images have afforded an opportunity to forecast yield as well as mapping spatial distribution on a near real-time basis for regions where this was previously not feasible, such as Africa . Nevertheless, at the present time satellite data is not yet mature enough to be used in traditional phenotyping platforms due to restricted pixel resolution. ...
Breeding is one of the central pillars of adaptation of crops to climate change. However, phenotyping is a key bottleneck that is limiting breeding efficiency. The awareness of phenotyping as a breeding limitation is not only sustained by the lack of adequate approaches, but also by the perception that phenotyping is an expensive activity. Phenotyping is not just dependent on the choice of appropriate traits and tools (e.g. sensors) but relies on how these tools are deployed on their carrying platforms, the speed and volume of data extraction and analysis (throughput), the handling of spatial variability and characterization of environmental conditions, and finally how all the information is integrated and processed. Affordable high throughput phenotyping aims to achieve reasonably priced solutions for all the components comprising the phenotyping pipeline. This mini-review will cover current and imminent solutions for all these components, from the increasing use of conventional digital RGB cameras, within the category of sensors, to open-access cloud-structured data processing and the use of smartphones. Emphasis will be placed on field phenotyping, which is really the main application for day-to-day phenotyping.
Maize has been identified as a strategic commodity for the reduction of poverty and enhancement of food security in the African continent. Climate variability and difficult economic conditions are pressuring farmers to produce higher (maize) yields with fewer inputs, per hectare. The remote sensing of crop specific structural parameters are essential in identifying the particular growth stages of the maize crop which require specific tasks from the farmer (e.g. weed control, top dressing, pesticide application for disease and borer control and critical moisture phase). This study sought to assess the performance of multiple linear regression (LR), Random Forest (RF) and Gaussian Process Regression (GPR) in the estimation of four maize crop structural parameters in a study area in the Vereeniging region of the Maize Triangle of South Africa. These parameters were leaf area index (LAI), stem height (HT), stem diameter (DIA) and stem density (SD). An additional aim was to investigate whether the combination of selected spectral vegetation indices (red-edge, chlorophyll, senescence and greenness) with Sentinel-2 reflectance bands as modelling predictors yielded improved results over the individual spectral bands alone. Combining reflectance bands and vegetation indices as modelling predictors yielded the highest validation accuracy, over other scenarios, for only one out of the four crop structural parameters (DAI). The reflectance bands only scenario yielded the highest validation accuracies for two crop structural parameters (HT and SD). The use of spectral vegetation indices alone as modelling predictors yielded the highest modelling accuracies for the LAI crop parameter than the other scenarios. These trends indicate that the combination of Sentinel-2 reflectance bands and derived vegetation indices do not always yield improved modelling results for the four crop structural parameters under investigation. As a result, reflectance bands (mostly) or indices alone could suffice for nearly all of the parameters. With respect to the modelling algorithms, LR yielded the highest accuracies for DIA and SD (Standard Error of Prediction or SEP values of 22.40%±4.65 and 34.15%±2.72 respectively). GPR yielded the highest accuracies for LAI and HT (SEP values of 28.69%±3.84 and 23.19%±2.27 respectively) while RF did not yield the highest validation accuracy for any of the crop structural parameters.
Imaging spectroscopy, also known as hyperspectral remote sensing, is based on the characterization of Earth surface materials and processes through spectrally-resolved measurements of the light interacting with matter. The potential of imaging spectroscopy for Earth remote sensing has been demonstrated since the 1980s. However, most of the developments and applications in imaging spectroscopy have largely relied on airborne spectrometers, as the amount and quality of space-based imaging spectroscopy data remain relatively low to date. The upcoming Environmental Mapping and Analysis Program (EnMAP) German imaging spectroscopy mission is intended to fill this gap. An overview of the main characteristics and current status of the mission is provided in this contribution. The core payload of EnMAP consists of a dual-spectrometer instrument measuring in the optical spectral range between 420 and 2450 nm with a spectral sampling distance varying between 5 and 12 nm and a reference signal-to-noise ratio of 400:1 in the visible and near-infrared and 180:1 in the shortwave-infrared parts of the spectrum. EnMAP images will cover a 30 km-wide area in the across-track direction with a ground sampling distance of 30 m. An across-track tilted observation capability will enable a target revisit time of up to four days at the Equator and better at high latitudes. EnMAP will contribute to the development and exploitation of spaceborne imaging spectroscopy applications by making high-quality data freely available to scientific users worldwide.
Maize crop production is constrained worldwide by nitrogen (N) availability and particularly in poor tropical and subtropical soils. The development of affordable high-throughput crop monitoring and phenotyping techniques is key to improving maize cultivation under low-N fertilization. In this study several vegetation indices (VIs) derived from Red-Green-Blue (RGB) digital images at the leaf and canopy levels are proposed as low-cost tools for plant breeding and fertilization management. They were compared with the performance of the normalized difference vegetation index (NDVI) measured at ground level and from an aerial platform, as well as with leaf chlorophyll content (LCC) and other leaf composition and structural parameters. A set of ten hybrids grown under five different nitrogen regimes were tested at the CIMMYT station of Harare (Zimbabwe). Grain yield and leaf N concentration across N fertilization levels were strongly predicted by most of these RGB indices (with R2 ~ 0.7), out performing the prediction power of the NDVI and LCC. RGB indices also outperformed the NDVI when assessing genotypic differences in grain yield and leaf N concentration within a given level of N fertilization. The best predictor of leaf N concentration across the five N regimes was LCC but its performance within treatments was inefficient. The leaf traits evaluated also seemed inefficient as phenotyping parameters. It is concluded that the adoption of RGB-based phenotyping techniques may significantly contribute to the progress of plant breeding and the appropriate management of fertilization.
With advances in satellite, airborne and ground based remote sensing, reflectance data are increasingly being used in agriculture. This paper reviews various remote sensing methods designed to optimize profitability of agricultural crop production and protect the environment. The paper presents examples of the use of remote sensing data in crop yield forecasting, assessing nutritional requirements of plants and nutrient content in soil, determining plant water demand and weed control.
Woody vegetation encroachment into grasslands or bush thickening, a global phenomenon, is transforming the Southern African grassland systems into savanna-like landscapes. Estimation of woody vegetation is important to rangeland scientists and land managers for assessing its impact on grass production and calculating its grazing and browsing capacity. Assessment of grazing and browsing components is often challenging because agro-ecological landscapes of this region are largely characterized by small scale and heterogeneous land-use-land-cover patterns. In this study, we investigated the utility of high spatial resolution remotely sensing data for modelling grazing and browsing capacity at landscape level. Woody tree density or Tree Equivalents (TE) and Total Leaf Mass (LMASS) data were derived using the Biomass Estimation for Canopy Volume (BECVOL) program. The Random Forest (RF) regression algorithm was assessed to establish relationships between these variables and vegetation indices (Simple Ratio and Normalized Difference Vegetation Index), calculated using the red and near infrared bands of SPOT5. The RF analysis predicted LMASS with R2 = 0.63 and a Root Mean Square Error (RMSE) of 1256 kg/ha compared to a mean of 2291kg/ha. TE was predicted with R2 = 0.55 and a RMSE = 1614 TE/ha compared to a mean of 3746 TE/ha. Next, spatial distribution maps of LMASS/ha and TE/ha were derived using separate RF regression models. The resultant maps were then used as input data into conventional grazing and browsing capacity models to calculate grazing and browsing capacity maps for the study area. This study provides a sound platform for integrating currently available and future remote sensing satellite data into rangeland carrying capacity modelling and monitoring.
Crop yield gap (Yg) can be disaggregated into two components: (i) one that is consistent across years and is, therefore, attributable to persistent factors that limit yields, and (ii) a second that varies from year to year due to inconsistent constraints on yields. Quantifying relative contributions of persistent and non-persistent factors to overall Yg, and identifying their underpinning causes, can help identify sound interventions to narrow current Yg and estimate magnitude of likely impact. The objective of this study was to apply this analytical framework to quantify the contribution of persistent factors to current Yg in high-yield irrigated maize systems in western US Corn Belt and identify some of the underpinning explanatory factors. We used a database containing producer yields collected during 10 years (2004–2013) from ca. 3000 irrigated fields in three regions of the state of Nebraska (USA). Yield potential was estimated for each region-year using a crop simulation model and actual weather and management data. Yg was calculated for each field-year as the difference between simulated yield potential and field yield. Two independent sources of field yield data were used: (i) producer-reported yields, and (ii) estimated yields using a combined satellite-crop model approach that does not rely on actual yield data. In each year (hereafter called ‘ranking years’), fields were grouped into ‘small’ and ‘large’ Yg categories. For a given category, Yg persistence was calculated by comparing mean Yg estimated for ranking years against mean Yg calculated, for the same group of fields, for the rest of the years. Explanatory factors for persistent Yg were assessed. Yg persistence ranged between ca. 30% and 50% across regions, with higher persistence in regions with heterogeneous soils. Estimates of Yg size and persistence based on producer-reported yields and satellite-model approach were in reasonable agreement, though the latter approach consistently underestimated Yg size and persistence. Small Yg category exhibited a higher frequency of fields with favorable soils and soybean-maize rotation and greater N fertilizer and irrigation inputs relative to the large Yg category. Remarkably, despite higher applied inputs, efficiencies in the use of N fertilizer, irrigation, and solar radiation were much higher in fields exhibiting small Yg. The framework implemented in this study can be applied to any cropping system for which a reasonable number of field-year yield and management data are available.
Traditional smallholder farming systems dominate the savanna range countries of sub-Saharan Africa and provide the foundation for the region’s food security. Despite continued expansion of smallholder farming into the surrounding savanna landscapes, food insecurity in the region persists. Central to the monitoring of food security in these countries, and to understanding the processes behind it, are reliable, high-quality datasets of cultivated land. Remote sensing has been frequently used for this purpose but distinguishing crops under certain stages of growth from savanna woodlands has remained a major challenge. Yet, crop production in dryland ecosystems is most vulnerable to seasonal climate variability, amplifying the need for high quality products showing the distribution and extent of cropland. The key objective in this analysis is the development of a classification protocol for African savanna landscapes, emphasizing the delineation of cropland. We integrate remote sensing techniques with probabilistic modeling into an innovative workflow. We present summary results for this methodology applied to a land cover classification of Zambia’s Southern Province. Five primary land cover categories are classified for the study area, producing an overall map accuracy of 88.18%. Omission error within the cropland class is 12.11% and commission error 9.76%
Validating land-cover maps at the global scale is a significant challenge. We built a global validation data-set based on interpreting Landsat Thematic Mapper (TM) and Enhanced TM Plus (ETM+) images for a total of 38,664 sample units pre-determined with an equal-area stratified sampling scheme. This was supplemented by MODIS enhanced vegetation index (EVI) time series data and other high-resolution imagery on Google Earth. Initially designed for validating 30 m-resolution global land-cover maps in the Finer Resolution Observation and Monitoring of Global Land Cover (FROM-GLC) project, the data-set has been carefully improved through several rounds of interpretation and verification by different image interpreters, and checked by one quality controller. Independent test interpretation indicated that the quality control correctness level reached 90% at level 1 classes using selected interpretation keys from various parts of the USA. Fifty-nine per cent of the samples have been verified with high-resolution images on Google Earth. Uncertainty in interpretation was measured by the interpreter's perceived confidence. Only less than 7% of the sample was perceived as low confidence at level 1 by interpreters. Nearly 42% of the sample units located within a homogeneous area could be applied to validating global land-cover maps whose resolution is 500 m or finer. Forty-six per cent of the sample whose EVI values are high or with little seasonal variation throughout the year can be applied to validate land-cover products produced from data acquired in different phenological stages, while approximately 76% of the remaining sample whose EVI values have obvious seasonal variation was interpreted from images acquired within the growing season. While the improvement is under way, some of the homogeneous sample units in the data-set have already been used in assessing other classification results or as training data for land-cover mapping with coarser-resolution data.
Agricultural development is an essential engine of growth and poverty reduction, yet agricultural data suffer from poor quality and narrow sectoral focus. There are several reasons for this: (1) difficult-to-measure smallholder agriculture is prevalent in poor countries; (2) agricultural data are collected with little coordination across sectors; and (3) poor analysis undermines the demand for high-quality data. This article argues that initiatives like the Global Strategy to Improve Agricultural and Rural Statistics bode well for the future. Moving from Devarajan’s statistical tragedy’ to Kiregyera’s statistical ‘renaissance’ will take a continued long-term effort by individual countries and development partners.
Climate change has the potential to significantly impact agricultural production and the stakes are large. Against this backdrop, two schools of thought find different projected impacts from climate change. On the one hand, crop models, based on plant physiology and developed and refined from field experiments over many decades, usually predict modestly negative to positive impacts from projected warming and rising carbon dioxide concentrations, both globally and in the US. The temperature distribution accounts for variation in temperatures across locations, using fine-scale weather data, and within days by assuming temperatures follow a sine curve between each day's maximum and minimum temperatures. Vapor pressure deficit (VPD) is calculated as the difference between how much water the air can hold when it is saturated and how much water it currently holds. The first two columns of the table display the marginal impact of each variable, which is assumed to be constant due to the linear functional form.
Lack of accurate maps on the extent of global cropland, and particularly the spatial distribution of major crop types, hampers policy and strategic investment and could potentially impede efforts to improve food security in an environment characterized by continued market volatility and a changing climate. Here we discuss the pressing need for the provision of spatially explicit cropland datasets at a global scale and review the strengths and weaknesses of the various approaches used to develop such data.
International Journal of Remote Sensing Publication details, including instructions for authors and subscription information: To cite this article: Oupa Malahlela, Moses Azong Cho & Onisimo Mutanga (2014) Mapping canopy gaps in an indigenous subtropical coastal forest using high-resolution WorldView-2 data, makes every effort to ensure the accuracy of all the information (the "Content") contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Accurate cropland maps at the global and local scales are crucial for scientists, government and nongovernment agencies, farmers and other stakeholders, particularly in food-insecure regions, such as Sub-Saharan Africa. In this study, we aim to qualify the crop classes of the MODIS Land Cover Product (LCP) in Sub-Saharan Africa using FAO (Food and Agricultural Organisation) and AGRHYMET (AGRiculture, Hydrology and METeorology) statistical data of agriculture and a sample of 55 very-high-resolution images. In terms of cropland acreage and dynamics, we found that the correlation between the statistical data and MODIS LCP decreases when we localize the spatial scale (from R 2 = 0.86 *** at the national scale to R 2 = 0.26 *** at two levels below the national scale). In terms of the cropland spatial distribution, our findings indicate a strong relationship between the user accuracy and the fragmentation of the agricultural landscape, as measured by the MODIS LCP; the accuracy decreases as the crop fraction increases. In addition, thanks to the Pareto boundary method, we were able to isolate and quantify the part of the MODIS classification error that could be directly linked to the performance of the adopted classification algorithm. Finally, based on these results, (i) a regional map of the MODIS LCP user accuracy estimates for cropland classes was produced for the entire Sub-Saharan region; this map presents a better accuracy in the western part of the region (43%–70%) compared to the eastern part (17%–43%); (ii) Theoretical user and producer accuracies for OPEN ACCESS Remote Sens. 2014, 6 8542 a given set of spatial resolutions were provided; the simulated future Sentinel-2 system would provide theoretical 99% user and producer accuracies given the landscape pattern of the region.
The quantification of aboveground biomass using remote sensing is critical for better understanding the role of forests in carbon sequestration and for informed sustainable management. Although remote sensing techniques have been proven useful in assessing forest biomass in general, more is required to investigate their capabilities in predicting intra-and-inter species biomass which are mainly characterised by non-linear relationships. In this study, we tested two machine learning algorithms, Stochastic Gradient Boosting (SGB) and Random Forest (RF) regression trees to predict intra-and-inter species biomass using high resolution RapidEye reflectance bands as well as the derived vegetation indices in a commercial plantation. The results showed that the SGB algorithm yielded the best performance for intra-and-inter species biomass prediction; using all the predictor variables as well as based on the most important selected variables. For example using the most important variables the algorithm produced an R2 of 0.80 and RMSE of 16.93 t·ha-1 for E. grandis; R2 of 0.79, RMSE of 17.27 t·ha-1 for P. taeda and R2 of 0.61, RMSE of 43.39 t·ha-1 for the combined species data sets. Comparatively, RF yielded plausible results only for E. dunii (R2 of 0.79; RMSE of 7.18 t·ha-1). We demonstrated that although the two statistical methods were able to predict biomass accurately, RF produced weaker results as compared to SGB when applied to combined species dataset. The result underscores the relevance of stochastic models in predicting biomass drawn from different species and genera using the new generation high resolution RapidEye sensor with strategically positioned bands.
Hyperspectral narrow-band (or imaging spectroscopy) spectral data are fast emerging as practical solutions in modeling and mapping vegetation. Recent research has demonstrated the advances in and merit of hyperspectral data in a range of applications including quantifying agricultural crops, modeling forest canopy biochemical properties, detecting crop stress and disease, mapping leaf chlorophyll content as it influences crop production, identifying plants affected by contaminants such as arsenic, demonstrating sensitivity to plant nitrogen content, classifying vegetation species and type, characterizing wetlands, and mapping invasive species. The need for significant improvements in quantifying, modeling, and mapping plant chemical, physical, and water properties is more critical than ever before to reduce uncertainties in our understanding of the Earth and to better sustain it. There is also a need for a synthesis of the vast knowledge spread throughout the literature from more than 40 years of research.
Hyperspectral Remote Sensing of Vegetation integrates this knowledge, guiding readers to harness the capabilities of the most recent advances in applying hyperspectral remote sensing technology to the study of terrestrial vegetation. Taking a practical approach to a complex subject, the book demonstrates the experience, utility, methods and models used in studying vegetation using hyperspectral data. Written by leading experts, including pioneers in the field, each chapter presents specific applications, reviews existing state-of-the-art knowledge, highlights the advances made, and provides guidance for the appropriate use of hyperspectral data in the study of vegetation as well as its numerous applications, such as crop yield modeling, crop and vegetation biophysical and biochemical property characterization, and crop moisture assessment.
This comprehensive book brings together the best global expertise on hyperspectral remote sensing of agriculture, crop water use, plant species detection, vegetation classification, biophysical and biochemical modeling, crop productivity and water productivity mapping, and modeling. It provides the pertinent facts, synthesizing findings so that readers can get the correct picture on issues such as the best wavebands for their practical applications, methods of analysis using whole spectra, hyperspectral vegetation indices targeted to study specific biophysical and biochemical quantities, and methods for detecting parameters such as crop moisture variability, chlorophyll content, and stress levels. A collective "knowledge bank," it guides professionals to adopt the best practices for their own work.
This paper presents a compendium of satellites under civilian and/or commercial control with the potential to gather global land-cover observations. From this we show that a growing number of sovereign states are acquiring capacity for space based land-cover observations and show how geopolitical patterns of ownership are changing. We discuss how the number of satellites flying at any time has progressed as a function of increased launch rates and mission longevity, and how the spatial resolutions of the data they collect has evolved. The first such satellite was launched by the USA in 1972. Since then government and/or private entities in 33 other sovereign states and geopolitical groups have chosen to finance such missions and 197 individual satellites with a global land-cover observing capacity have been successfully launched. Of these 98 were still operating at the end of 2013. Since the 1970s the number of such missions failing within 3 years of launch has dropped from around 60% to less than 20%, the average operational life of a mission has almost tripled, increasing from 3.3 years in the 1970s to 8.6 years (and still lengthening), the average number of satellites launched per-year/per-decade has increased from 2 to 12 and spatial resolution increased from around 80 m to less than 1 m multispectral and less than half a meter for panchromatic; synthetic aperture radar resolution has also fallen, from 25 m in the 1970s to 1 m post 2007. More people in more countries have access to data from global land-cover observing spaceborne missions at a greater range of spatial resolutions than ever before. We provide a compendium of such missions, analyze the changes and shows how innovation, the need for secure data-supply, national pride, falling costs and technological advances may underpin the trends we document.
The visual progression of sirex (Sirex noctilio) infestation symptoms has been categorized into three distinct infestation phases, namely the green, red and grey stages. The grey stage is the final stage which leads to almost complete defoliation resulting in dead standing trees or snags. Dead standing pine trees however, could also be due to the lightning damage. Hence, the objective of the present study was to distinguish amongst healthy, sirex grey-attacked and lightning-damaged pine trees using AISA Eagle hyperspectral data, random forest (RF) and support vector machines (SVM) classifiers. Our study also presents an opportunity to look at the possibility of separating amongst the previously mentioned pine trees damage classes and other landscape classes on the study area. The results of the present study revealed the robustness of the two machine learning classifiers with an overall accuracy of 74.50% (total disagreement = 26%) for RF and 73.50% (total disagreement = 27%) for SVM using all the remaining AISA Eagle spectral bands after removing the noisy ones. When the most useful spectral bands as measured by RF were exploited, the overall accuracy was considerably improved; 78% (total disagreement = 22%) for RF and 76.50% (total disagreement = 24%) for SVM. There was no significant difference between the performances of the two classifiers as demonstrated by the results of McNemar’s test (chi-squared; χ2 = 0.14, and 0.03 when all the remaining ASIA Eagle wavebands, after removing the noisy ones and the most important wavebands were used, respectively). This study concludes that AISA Eagle data classified using RF and SVM algorithms provide relatively accurate information that is important to the forest industry for making informed decision regarding pine plantations health protocols.
This experiment investigated the relationship between tobacco canopy spectral characteristics and tobacco biomass. A completely randomized design, with plantings on the 15th of September, October, November, and December, each with 9 variety × fertiliser management treatments, was used. Starting from 6 weeks after planting, reflectance measurements were taken from one row, using a multispectral radiometer. Individual plants from the other 3 rows were also measured, and the above ground whole plants were harvested and dried for reflectance/dry mass regression analysis. The central row was harvested, cured, and weighed. Both the maximum NDVI and mass at untying declined with later planting and so was the mass-NDVI coefficient of determination. The best fitting curves for the yield-NDVI correlations were quadratic. September reflectance values from the October crop reflectance were statistically similar (), while those for the November and the December crops were significantly different () from the former two. Mass at untying and NDVI showed a quadratic relationship in all the three tested varieties. The optimum stage for collecting spectral data for tobacco yield estimation was the 8–12 weeks after planting. The results could be useful in accurate monitoring of crop development patterns for yield forecasting purposes.
Previous studies have shown the importance of soil moisture (SM) in estimating crop yield potential (YP). The sensor based nitrogen (N) rate calculator (SBNRC) developed by Oklahoma State University utilizes the Normalized Difference Vegetation Index (NDVI) and the in-season estimated yield (INSEY) as the estimate of biomass to assess YP and to generate N recommendations based on estimated crop need. The objective was to investigate whether including the SM parameter into SBNRC could help to increase the accuracy of YP prediction and improve N rate recommendations. Two experimental sites (Lahoma and Perkins) in Oklahoma were established in 2006/07 and 2007/08. Wheat spectral reflectance was measured using a GreenSeeker™ 505 hand-held optical sensor (N-Tech Industries, Ukiah, CA). Soil–water content measured with matric potential 229-L sensors (Campbell Scientific, Logan, UT) was used to determine volumetric water content and fractional water index. The relationships between NDVI, INSEY and SM indices at planting and sensing at 5, 25, 60 and 75-cm depths versus grain yield (GY) were evaluated. Wheat GY, NDVI at Feekes 5 and soil WC at planting and as sensed at three depths were also analyzed for eight consecutive growing seasons (1999–2006) for Lahoma. Incorporation of SM into NDVI and INSEY calculations resulted in equally good prediction of wheat GY for all site-years. This indicates that NDVI alone was able to account for the lack of SM information and thus lower crop YP. Soil moisture data, especially at the time of sensing at the 5-cm depth could assist in refining winter wheat YP prediction.
Field experiments and simulation models are useful tools for understanding crop yield gaps, but scaling up these approaches to understand entire regions over time has remained a considerable challenge. Satellite data have repeatedly been shown to provide information that, by themselves or in combination with other data and models, can accurately measure crop yields in farmers’ fields. The resulting yield maps provide a unique opportunity to overcome both spatial and temporal scaling challenges and thus improve understanding of crop yield gaps. This review discusses the use of remote sensing to measure the magnitude and causes of yield gaps. Examples from previous work demonstrate the utility of remote sensing, but many areas of possible application remain unexplored. Two simple yet useful approaches are presented that measure the persistence of yield differences between fields, which in combination with maps of average yields can be used to direct further study of specific factors. Whereas the use of remote sensing may have historically been restricted by the cost and availability of fine resolution data, this impediment is rapidly receding.
The large-scale monitoring and estimation of crop yield is essential for food security in Mexico. This study developed and validated a method of monitoring and estimating corn (Zea mays L.) yield by means of satellite and ground-based data. In autumn-winter 1999 and spring-summer 2000, eight locations under irrigated and nonirrigated conditions in corn valleys of Mexico were localized by Global Positioning Systems (GPS) and were sampled every 15 d. Photosynthetic active radiation (PAR), leaf area index (LAI), crop development stage (DVS), planting dates, and grain yield data were gathered from the field. The normalized difference vegetation index (NDVI) was derived from NOAA-Advanced Very High Resolution Radiometer (AVHRR) images. A growth model was developed to integrate satellite and ground data. Net primary productivity (NPP) was estimated using PAR and NDVI. Dry weight increase (kg ha-1 d-1) was determined considering NPP and the partitioning factor. Results indicated that the model accounts for 89% of the variability in yields under irrigated conditions and 76% under nonirrigated conditions. The methodology seems advantageous in large-scale monitoring and assessment of corn yield.
The emergence of satellite sensors that can routinely observe millions of individual smallholder farms raises possibilities for monitoring and understanding agricultural productivity in many regions of the world. Here we demonstrate the potential to track smallholder maize yield variation in western Kenya, using a combination of 1-m Terra Bella imagery and intensive field sampling on thousands of fields over 2 y. We find that agreement between satellite-based and traditional field survey-based yield estimates depends significantly on the quality of the field-based measures, with agreement highest ([Formula: see text] up to 0.4) when using precise field measures of plot area and when using larger fields for which rounding errors are smaller. We further show that satellite-based measures are able to detect positive yield responses to fertilizer and hybrid seed inputs and that the inferred responses are statistically indistinguishable from estimates based on survey-based yields. These results suggest that high-resolution satellite imagery can be used to make predictions of smallholder agricultural productivity that are roughly as accurate as the survey-based measures traditionally used in research and policy applications, and they indicate a substantial near-term potential to quickly generate useful datasets on productivity in smallholder systems, even with minimal or no field training data. Such datasets could rapidly accelerate learning about which interventions in smallholder systems have the most positive impact, thus enabling more rapid transformation of rural livelihoods.
The application of remote sensing has progressed remarkably over
the years, with technological advancements that have led to the
availability of efficient, relatively cheap, robust, as well as high
resolution images. Remarkable progress has been made on crop,
forest and rangeland monitoring, using satellite data. This paper
reviews progress in the application of remote sensing technologies
in South Africa with a specific focus on vegetation monitoring.
Vegetation state monitoring has been identified by the South African
Mission Advisory Committee (EO – MAC) as the primary objective of
the proposed EO-SAT1 satellite to be launched in the near future. The
paper provides a review of the developments in the science of satellite
remote sensing as it quantifies the studies that have been done, using
specific sensors and evaluates the importance of sensor resolution
and data availability for South Africa. Specific application examples
are used to showcase the developments. An analysis of studies shows
that South African scientists have used a wide range of satellite data
and developed novel processing techniques with applications specific
to the environmental conditions of the region. However, this work has
observed that, for accurate and reliable vegetation monitoring at a
national scale, access to relatively cheap satellite data with optimal
spectral band information is limited.
Equations are presented to estimate production from multi-temporal sums of AVHRR vegetation indices, together with the associated errors. The underlying remote sensing and primary production models are discussed and a physically based model is tested using a data set for Mali for which atmospherically corrected and sensor calibrated data are available. The possible interference of mixtures of surface brightnesses in the scene on satellite measurements of vegetation indices is discussed. Application of the model to derive biological parameters such as the efficiency of conversion of solar radiation into plant material is demonstrated. -from Author
Accurate, reliable, and up-to-date forest stand volume information is a prerequisite for a detailed evaluation of commercial forest resources and their sustainable management. Commercial forest responses to global climate change remain uncertain, and hence the mapping of stand volume as carbon sinks is fundamentally important in understanding the role of forests in stabilizing climate change effects. The aim of this study was to examine the utility of stochastic gradient boosting (SGB) and multi-source data to predict stand volume of a Eucalyptus plantation in South Africa. The SGB ensemble, random forest (RF), and stepwise multiple-linear regression (SMLR) were used to predict Eucalyptus stand volume and other related tree-structural attributes such as mean tree height and mean diameter at breast height (DBH). Multi-source data consisted of SPOT-5 raw spectral features (four bands), 14 spectral vegetation indices, rainfall data, and stand age. When all variables were used, the SGB algorithm showed that stand volume can be accurately estimated (R2 = 0.78 and RMSE = 33.16 m3 ha−1 (23.01% of the mean)). The competing RF ensemble produced an R2 value of 0.76 and a RMSE value of 37.28 m3 ha−1 (38.28% of the mean). SMLR on the other hand, produced an R2 value of 0.65 and an RMSE value of 42.50 m3 ha−1 (42.50% of the mean). Our study further showed that Eucalyptus mean tree height (R2 = 0.83 and RMSE = 1.63 m (9.08% of the mean)) and mean diameter at breast height (R2 = 0.74 and RMSE = 1.06 (7.89% of the mean)) can also be reasonably predicted using SGB and multi-source data. Furthermore, when the most important SGB model-selected variables were used for prediction, the predictive accuracies improved significantly for mean DBH (R2 = 0.81 and RMSE = 1.21 cm (6.12% of the mean)), mean tree height (R2 = 0.86 and RMSE = 1.39 m (7.02% of the mean)), and stand volume (R2 = 0.83 and RMSE = 29.58 m3 ha−1 (17.63% of the mean)). These results underscore the importance of integrating multi-source data with remotely sensed data for predicting Eucalyptus stand volume and related tree-structural attributes.
Airborne remote-sensing has been identified worldwide as a promising technique for identifying and mapping weeds in crops, and potentially offers a solution to the current logjam in precision weed management: namely, the ability to generate timely and accurate weed maps. One of the main advantages of remote-sensing is that synoptic weed data can be acquired virtually instantaneously (within the field of view of the sensor), and a weed map generated within hours of data acquisition. However, because little information is available concerning the scale at which weeds should be managed within fields, the sensing and mapping technology has tended to dictate the resolution at which weeds must be mapped. This paper summarizes the work completed to date to investigate the use of airborne remote-sensing for weed mapping in crops, and discusses application of the technology in precision weed management practices.
Monitoring crop condition and production estimates at the state and county level is of great interest to the U.S. Department of Agriculture. The National Agricultural Statistical Service (NASS) of the U.S. Department of Agriculture conducts field interviews with sampled farm operators and obtains crop cuttings to make crop yield estimates at regional and state levels. NASS needs supplemental spatial data that provides timely information on crop condition and potential yields. In this research, the crop model EPIC (Erosion Productivity Impact Calculator) was adapted for simulations at regional scales. Satellite remotely sensed data provide a real-time assessment of the magnitude and variation of crop condition parameters, and this study investigates the use of these parameters as an input to a crop growth model. This investigation was conducted in the semi-arid region of North Dakota in the southeastern part of the state. The primary objective was to evaluate a method of integrating parameters retrieved from satellite imagery in a crop growth model to simulate spring wheat yields at the sub-county and county levels. The input parameters derived from remotely sensed data provided spatial integrity, as well as a real-time calibration of model simulated parameters during the season, to ensure that the modeled and observed conditions agree. A radiative transfer model, SAIL (Scattered by Arbitrary Inclined Leaves), provided the link between the satellite data and crop model. The model parameters were simulated in a geographic information system grid, which was the platform for aggregating yields at local and regional scales. A model calibration was performed to initialize the model parameters. This calibration was performed using Landsat data over three southeast counties in North Dakota. The model was then used to simulate crop yields for the state of North Dakota with inputs derived from NOAA AVHRR data. The calibration and the state level simulations are compared with spring wheat yields reported by NASS objective yield surveys.
Regional crop yield prediction is a significant component of national food policy making and security assessments. A data assimilation method that combines crop growth models with remotely sensed data has been proven to be the most effective method for regional yield estimates. This paper describes an assimilation method that integrates a time series of leaf area index (LAI) retrieved from ETM+ data and a coupled hydrology-crop growth model which links a crop growth model World Food Study (WOFOST) and a hydrology model HYDRUS-1D for regional maize yield estimates using the ensemble Kalman filter (EnKF). The coupled hydrology-crop growth model was calibrated and validated using field data to ensure that the model accurately simulated associated state variables and maize growing processes. To identify the parameters that most affected model output, an extended Fourier amplitude sensitivity test (EFAST) was applied to the model before calibration. The calibration results indicated that the coupled hydrology-crop growth model accurately simulated maize growth processes for the local cultivation variety tested. The coefficient of variations (CVs) for LAI, total above-ground production (TAGP), dry weight of storage organs (WSO), and evapotranspiration (ET) were 13%, 6.9%, 11% and 20%, respectively. The calibrated growth model was then combined with the regional ETM+ LAI data using a sequential data assimilation algorithm (EnKF) to incorporate spatial heterogeneity in maize growth into the coupled hydrology-crop growth model. The theoretical LAI profile for the near future and the final yield were obtained through the EnKF algorithm for 50 sample plots. The CV of the regional yield estimates for these sample plots was 8.7%. Finally, the maize yield distribution for the Zhangye Oasis was obtained as a case study. In general, this research and associated model could be used to evaluate the impacts of irrigation, fertilizer and field management on crop yield at a regional scale.
The prospect of regular assessments of insect defoliation using remote sensing technologies has increased in recent years through advances in the understanding of the spectral reflectance properties of vegetation. The aim of the present study was to evaluate the ability of the red edge channel of Rapideye imagery to discriminate different levels of insect defoliation in an African savanna by comparing the results of obtained from two classifiers. Random Forest and Support vector machine classification algorithms were applied using different sets of spectral analysis involving the red edge band. Results show that the integration of information from red edge increases classification accuracy of insect defoliation levels in all analysis performed in the study. For instance, when all the 5 bands of Rapideye imagery were used for classification, the overall accuracies increases about 19% and 21% for SVM and RF, respectively, as opposed to when the red edge channel was excluded. We also found out that the normalized difference red-edge index yielded a better accuracy result than normalized difference vegetation index. We conclude that the red-edge channel of relatively affordable and readily available high-resolution multispectral satellite data such as Rapideye has the potential to considerably improve insect defoliation classification especially in sub-Saharan Africa where data availability is limited.
Drought is one of the major environmental disasters in southern Africa. In recent years, the damage from droughts to the environment and economies of some countries was extensive, and the death toll of livestock and wildlife was unprecedented. Weather data often come from a very sparse meteorological network, incomplete and/or not always available in good time to enable relatively accurate and timely large scale drought detection and monitoring. Therefore, data obtained from the Advanced Very High Resolution Radiometer (AVHRR) sensor on board the NOAA polar-orbiting satellites have been studied as a tool for drought monitoring and climate impact assessment in southern Africa. The AVHRR-based vegetation condition index (VCI) and temperature condition index (TCI) developed recently were used in this study because in other parts of the globe they showed good results when used for drought detection and tracking, monitoring excessive soil wetness, assessment of weather impacts on vegetation, and evaluation of vegetation health and productivity. The results clearly show that temporal and spatial characteristics of drought in southern Africa can be detected, tracked, and mapped by the VCI and TCI indices. These results were numerically validated by in situ data such as precipitation, atmospheric anomaly fields, and agricultural crop yield. In the later case, it was found that usable corn yield scenarios can be constructed from the VCI and TCI at approximately 6 (in some regions up to 13) weeks prior to harvest time. These indices can be especially beneficial when used together with ground data.
This study integrated environmental variables together with high spectral resolution WorldView-2 imagery to detect and map Thaumastocoris peregrinus damage in Eucalypt plantation forests in KwaZulu-Natal, South Africa. The WorldView-2 bands, vegetation indices and environmental variables were entered separately into PLS regression models to predict T. peregrinus damage. The datasets were then integrated to test the collective strength in predicting T. peregrinus damage. Important variables were identified by variable importance (VIP) scores and were re-entered into a PLS regression model. The VIP model was then extrapolated to map the severity of damage and predicted T. peregrinus damage with an R2 value of 0.71 and a RMSE of 3.26% on an independent test dataset. The red edge and near-infrared bands of the WorldView-2 sensor together with the temperature dataset were identified as important variables in predicting T. peregrinus damage. The results indicate the potential of integrating WorldView-2 data and environmental variables to improve the mapping and monitoring of insect outbreaks in plantation forests. The result is critical for plantation health monitoring using a new sensor which contains important vegetation wavelengths.
Four timely and broadly available remotely sensed datasets were assessed for inclusion into county-level corn and soybean yield forecasting efforts focused on the Corn Belt region of the central United States (US). Those datasets were the (1) Normalized Difference Vegetation Index (NDVI) as derived from the Terra satellite's Moderate Resolution Imaging Spectroradiometer (MODIS), (2) daytime and (3) nighttime land surface temperature (LST) as derived from Aqua satellite's MODIS, and (4) precipitation from the National Weather Service (NWS) Nexrad-based gridded data product. The originating MODIS data utilized were the globally produced 8-day, clear sky composited science products (MOD09Q1 and MYD11A2), while the US-wide NWS data were manipulated to mesh with the MODIS imagery both spatially and temporally by regridding and summing the otherwise daily measurements. The crop growing seasons of 2006–2011 were analyzed with each year bounded by 32 8-day periods from mid-February through late October. Land cover classifications known as the Cropland Data Layer as produced annually by the National Agricultural Statistics Service (NASS) were used to isolate the input dataset pixels as to corn and soybeans for each of the corresponding years. The relevant pixels were then averaged by crop and time period to produce a county-level estimate of NDVI, the LSTs, and precipitation. They in turn were related to official annual NASS county level yield statistics. For the Corn Belt region as a whole, both corn and soybean yields were found to be positively correlated with NDVI in the middle of the summer and negatively correlated to daytime LST at that same time. Nighttime LST and precipitation showed no correlations to yield, regardless of the time prior or during the growing season. There was also slight suggestion of low NDVI and high daytime LST in the spring being positively related to final yields, again for both crops. Taking only NDVI and daytime LST as inputs from the 2006–2011 dataset, regression tree-based models were built and county-level, within-sample coefficients of determination (R2) of 0.93 were found for both crops. Limiting the models by systematically removing late season data showed the model performance to remain strong even at mid-season and still viable even earlier. Finally, the derived models were used to predict out-of-sample for the 2012 season, which ended up having an anomalous drought. Yet, the county-level results compared reasonably well against official statistics with R2 = 0.77 for corn and 0.71 for soybeans. The root-mean-square errors were 1.26 and 0.42 metric tons per hectare, respectively.
Thaumastocoris peregrinus is an insect that causes significant damage to Eucalyptus
plantations internationally. This bug inhibits the photosynthetic ability of the tree,
resulting in stunted growth and even death of severely infested trees. This study uses
high spatial resolution satellite imagery (WorldView-2 sensor data), with unique band
settings for the prediction of T. peregrinus damage in plantation forests using partial
least squares (PLS) regression. The PLS models developed from the WorldView-2 sensor
bands and indices were inverted to map the severity of the damage caused by
the pest. The WorldView-2 sensor bands and indices predicted T. peregrinus damage
with an R2 value of 0.65 and a root mean square error (RMSE) of 3.62% on an
independent test data set. The red-edge and near-infrared bands of the WorldView-2 sensor
and pigment-specific indices and red-edge indices were identified as significant
bands by variable importance scores for the prediction of T. peregrinus damage. This
study demonstrates the potential of WorldView-2 sensor data in successfully predicting
T. peregrinus damage using PLS regression and identifies important spectral variables
for the prediction of forest damage in plantation forests.
A leading text for undergraduate- and graduate-level courses, this book introduces widely used forms of remote sensing imagery and their applications in plant sciences, hydrology, earth sciences, and land use analysis. The text provides comprehensive coverage of principal topics and serves as a framework for organizing the vast amount of remote sensing information available on the Web. Featuring case studies and review questions, the book's 4 sections and 21 chapters are carefully designed as independent units that instructors can select from as needed for their courses. Illustrations include 29 color plates and over 400 black-and-white figures.
New to This Edition
Reflects significant technological and methodological advances.
Chapter on aerial photography now emphasizes digital rather than analog systems.
Updated discussions of accuracy assessment, multitemporal change detection, and digital preprocessing.
Links to recommended online videos and tutorials.
Recent investigations have demonstrated that inter-year NOAAAVHRR NDVI variations in the middle of the rainy season can provide information on final crop yield in Sahelian countries. The present work continues this line of research by the use of 10-day Global Area Coverage (GAC) NDVI Maximum Value Composites, which are widely available and cost-effective in Africa. This use actually posed some problems which were mitigated by a multistep methodology aimed at forecasting millet and sorghum yield in Niger. The soil effect was first minimized in the NDVI images, and a geographical standardization was applied to the sub-district mean NDVI values and to the relevant ground yield estimates in order to remove most of the noninteresting information related to variations in land resources. A correlation analysis on the data obtained showed that the best period for yield forecasting was from the end of August to the middle of September. A further improvement in the forecasting capability of the procedure was then achieved by an image-based statistical identification of the most intensively cultivated areas. The final result of the complete methodology was the forecast of crop yield within the middle of September with an acceptable level of accuracy (mean error of 72 kg ha).
Crop production assessments are extremely valuable because of their economic importance in influencing international trade and national economic policies. This study investigates using the Vegetation Condition Index derived from NOAA/AVHRR satellite data to estimate the maize production for the United States Corn Belt. Satellite data from 1985 to 1992 are utilized within a model and explain more than 50 per cent of the variation in the normalized yields from 42 Crop Reporting Districts. Results estimating the regional maize production are encouraging, and operational estimates using this model would be available for about two months prior to the maize harvest in the Corn Belt.
Successful application of the normalized difference vegetation index (NDVI) for estimating weather impacts on vegetation is currently hindered in non-homogeneous areas. The problem is that the differences between the level of vegetation in these areas can be related, in addition to weather impacts, to the differences in geographic resources (climate, soil, vegetation type and topography). These differences should be eliminated when weather impacts on vegetation are estimated from NDVI data. This paper discusses a concept and a technique for eliminating that portion of the NDVI which is related to the contribution of geographic resources to the amount of vegetation. The Advanced Very High Resolution Radiometer (AVHRR) data of the Global Vegetation Index format were used for the 1984-1987 seasons in Sudan. The procedure suggests normalization of NDVI values relative to the absolute maximum and the absolute minimum of NDVI. These two criteria were shown to be an appropriate characteristic of geographic resources of an area. The modified NDVI was named the Vegetation Condition Index (VCI). Comparison between VCI, NDVI and precipitation dynamics showed that the VCI estimates better portray precipitation dynamics as compared to the NDVI. The VCI permits not only the desciption of vegetation but also estimation of spatial and temporal vegetation changes and weather impacts on vegetation.
Spectral reflectances of several crops at Phoenix, Arizona were measured during two growing seasons using a hand-held radiometer, the Exotech Model 100A, that had a spectral bandpass configuration similar to scanning radiometers aboard Landsat 2 and 3. During the period of grain filling, yields of two wheat and one barley variety were well correlated with the integrated daily values of a modified vegetation index derived from reflectances in MSS Bands 5 and 7 (0.6-0.7 and 0.8-1.1 μm respectively). The derived model accounted for 88 per cent of the variability in yields from 103 to 656 g/m2 which were due to differential experimental soil moisture conditions (20 to 70 cm applied water).
A regression model approach using a normalized difference vegetation index (NDVI) has the potential for estimating crop production in East Africa. However, before production estimation can become a reality, the underlying model assumptions and statistical nature of the sample data (NDVI and crop production) must be examined rigorously. Annual maize production statistics from 1982-90 for 36 agricultural districts within Kenya were used as the dependent variable; median area NDVI (independent variable) values from each agricultural district and year were extracted from the annual maximum NDVI data set. The input data and the statistical association of NDVI with maize production for Kenya were tested systematically for the following items: (1) homogeneity of the data when pooling the sample, (2) gross data errors and influence points, (3) serial (time) correlation, (4) spatial autocorrelation and (5) stability of the regression coefficients. The results of using a simple regression model with NDVI as the only independent variable are encouraging (r 0.75, p 0.05) and illustrate that NDVI can be a responsive indicator of maize production, especially in areas of high NDVI spatial variability, which coincide with areas of production variability in Kenya.
On a 1984-1989 series of ARTEMIS-NDVI data derived from the NOAA-AVHRR sensor a case study on crop monitoring and early crop yield forecasting was elaborated for the provinces of Burkina Faso. In order to remove residual effects of clouds and other atmospheric influences on 10-day maximum NDVI images, a conditional temporal interpolation method was applied. Various NDVI regression parameters were compared. For the seven northern provinces, a simple linear regression based on averaged maximum 10-daily or monthly NDVI values proved to be superior to regressions based on the integrated NDVt and on NDVI increments. Multiple regressions led to significantly higher correlation coefficients, but only towards the end of the growing season (up to r2 = 087). The simple linear regression was also found valid for a part of the central and southern provinces. The yields of the majority of the provinces however was best approximated using one second-order polynomial equation. A test of the regressions on 1989 data showed a forecast error percentage of less than 15 per cent for half of the 30 provinces in August, approximately 2 months before harvest. In the other half of the provinces, high forecast errors occurred mainly due to a locust invasion, excessive rainfall in August and drought in September, after the time of the forecast. Therefore correction factors for the occurrence of extreme pest and other problems have to be included in the model in close cooperation with the relevant organizations. Some of these problems could however be assessed indirectly from the NDVI dynamics.