Assessing the Within-Field Heterogeneity Using Rapid-Eye NDVI Time Series Data

Abstract

(1) Background: The relation between the sub-field heterogeneity of soil properties and high-resolution satellite time series data might help to explain spatiotemporal patterns of crop growth, but detailed field studies are seldom. (2) Methods: Normalized Difference Vegetation Index (NDVI) data derived from satellite time series images were used to identify changes in the spatial distribution of winter triticale (×Triticosecale), winter rye (Secale cereale) and winter barley (Hordeum vulgare) growth (2015 to 2020) for a field in north-eastern Germany. NDVI patterns (quartiles) that remained persistent over time were identified and it was tested if spatially heterogeneous soil characteristics such as water holding capacity and altitude could explain them. (3) Results: A statistically significant relationship between elevation and soil classes with NDVI values was found in most cases. The lowest NDVI quartiles, considered as representing the poorest growth conditions, were generally found in the depressions with the lowest water holding capacity. These areas showed temporally stable spatial patterns, especially during the pre-harvest period. Over the six-year period, up to 80% of the grid cells with the lowest NDVI values were spatially consistent over time. Differences in the climatic water balance were rather low but could contribute to explaining spatial patterns, such as the lower clustering of values in the wettest year. (4) Conclusions: High-resolution satellite NDVI time series are a valuable information source for precise land management in order to optimize crop management with respect to yield and ecosystem services.

1. Introduction

Heterogeneous growing conditions within a field plot can benefit from modern precision agriculture in terms of spatially optimized management [1] and illustrate the high importance of acknowledging the interaction between the soil conditions, topography and cereal production [2,3,4,5,6,7]. For an efficient and non-destructive assessment of the heterogeneity of soil conditions of arable land and plant growth characteristics, the use of remote sensing data is becoming increasingly established [8,9]. Plant growth across landscapes or within a single field can show temporally consistent spatial patterns, such as low- and high-yielding areas. Depending on the spatial resolution, vegetation indices (VI) allow for assessing such patterns and their change over time. Several studies have successfully used vegetation indices based on remote sensing data to monitor and predict plant growth at the field scale [4,5,10,11,12]. However, the direct link between within-field changes in soil properties, topography and plant growth and their change over time has seldom been studied in more detail. It therefore often remains unclear if spatial growth differences strongly vary over a longer time period and are governed by external factors such as meteorological conditions, or are mainly influenced by intrinsic factors, such as soil parameters, which would lead to temporally consistent spatial growth patterns.

This study aims at providing an example and a workflow for using high-resolution satellite time series (Rapid Eye, Planet Scope) to assess spatially consistent vegetation patterns and their relation to soil characteristics at the field scale. The test site is an arable field of about six ha located in the state of Brandenburg, Germany. Using satellite data from the years 2015 to 2020 the following research hypotheses were tested: (1) the spatial distribution of Normalized Difference Vegetation Index (NDVI) mainly depends on soil heterogeneity and is consistent over time, (2) annual weather conditions can slightly modify the spatial pattern of NDVI. Years with dry conditions lead to more pronounced spatial patterns.

2. Materials and Methods

The study area is located in Großmutz, Brandenburg, Germany, between continental (low precipitation, low humidity) and maritime (high precipitation, high humidity) climates [13] and is classified as a moderately dry lowland climate. The long-term averages between 1991 and 2020 (WHO climate normal) of the closest climate station of the German Weather Service (DWD) in Neuruppin-Alt Rupin (Station ID: 96) are 623 mm of annual precipitation and 9.6 °C mean air temperature [14,15]. The field has an area of six ha and is characterized by spatial heterogeneity in terms of soil properties, topography and crop growth [16]. The altitude decreases from the southeastern corner (58 m a.s.l.) to the northern corner (51 m a.s.l.) (Figure 1). A soil reconnaissance map by [17] classifies the investigated area as an anhydromorphic substrate type “ASL w” (sandy ground moraine with undulated relief). The dominant soil types can be classified as Arenic Cambisol or Haplic Luvisol [18]. The agricultural field is conventionally managed and not irrigated. The cereal crops cultivated during the observation period are listed in

Table 1. Cultivated crops on the experimental site in chronological order. The growing season here refers to the time span between the sowing date in the first year and the harvest date in the subsequent year.

Growing Season	Specie	Cultivar
2014/15	Winter triticale (×Triticosecale)	Talentro
2015/16	Winter triticale (×Triticosecale)	Lombardo
2016/17	Winter rye (Secale cereale)	KWS Bono
2017/18	Summer oat (Avena sativa)	Poseidon
2018/19	Winter barley (Hordeum vulgare)	Anja
2019/20	Winter triticale (×Triticosecale)	Lombardo

. Summer oat (harvested in 2018) was excluded from this study to facilitate interannual comparisons between winter cereals.

The meteorological data used in this study were based on station data provided by the DWD (Germany’s national meteorological service (Deutscher Wetterdienst, DWD)) [18]. The climatic water balance (CWB in mm) was calculated for each day by subtracting the potential evaporation from the precipitation. Daily water balances were summed up to monthly values.

Soil samples were taken up to one m depth with a Pürckhauer soil auger along nine transect lines as described in [16]. The transect orientation followed the slope from the southwestern side to the northern corner with a spatial distance of 50 m from each other. The outer transects were 25 m apart from the edge of the field. Within the transects, samples were taken in 15 m intervals and each coordinate of the 87 sampling points was recorded with a Trimble Geo 7× Handheld GNSS System (Sunnyvale, CA, USA). Following the Munsell soil color charts [19] and the FAO (Food and Agriculture Organization of the United Nations) guidelines for soil description [20], the physical properties of the soil profiles were estimated in moist conditions by manually wetting soil probes. In general, all soil profiles down to 1 m depth were composed of a layer with sandy soil material (predominantly Loamy Sand or Sand) with variable thickness (35 to 100 cm) and a subsequent loam layer (Sandy loam to Loam) with variable thickness (0–65 cm). Based on the spatial variability of the vertical position (depth) of the loamy layer, three soil classes were derived as an indicator for differences in the available water capacity (AWC) of the soils [16]. Soil class 1 (S1) was characterized by a shallow upper boundary of the loamy subsoil layer (<60 cm depth), class 2 (S2), a deeper upper boundary of the loamy layer at 60–100 cm depth and class 3 (S3) was sandy throughout 100 cm soil depth (no loamy layer). Assuming that sandy soils have a low water holding capacity [21] that is improved by the presence of an underlying loamy layer, soil class 1 is assumed to show the highest and class 3 the lowest AWC. An interpolated soil map was created by kriging the soil sample points. To take the different altitude levels into account, which describe the position of the drilling points along the slopes (catena) at our site, a freely available digital terrain model (DTM) provided by the Amt für Landesvermessung und Geobasisinformation Brandenburg (LGB, https://geobasis-bb.de/lgb/de/geodaten/3d-produkte/gelaendemodell/, last visited: 17 October 2022) was used to further distinguish between three altitude classes: A1: <53.5 m, A2: 53.5–55.37 m, A3: >55.37 m a.s.l. Both classifications are shown in Figure 1.

The main type of satellite imagery used in this paper originates from the RapidEye satellite constellation. RapidEye was owned by Planet (Planet Labs, Inc., San Francisco, CA, USA) and the data were provided in cooperation with ESA’s Third-Party Missions Programme. The RapidEye satellite program consisted of 5 satellites collecting data in five spectral bands: red: 630–685 nm, green: 520–590 nm, blue: 440–510 nm, red edge: 690–730 nm and near-infrared (NIR): 760–850 nm. The resolution was 5 × 5 m per grid cell on orthorectified imagery resulting from 6.5 m GSD (Ground Sampling Distance) at the nadir [22]. Only images captured within the main growing season after winter were selected, which was between April and June of each year. Cloud coverage was visually checked. Since RapidEye’s operation was discontinued in spring 2020, PlanetScope imagery was applied as the satellite data source for the 2020 data. PlanetScope was chosen because the images were delivered by the same provider (Planet Labs, Inc., San Francisco, CA, USA), had a similar high temporal and spatial resolution compared to RapidEye and were available free of charge, including an atmospheric correction. The first 12 satellites were launched on 22 June 2016; therefore, PlanetScope was not able to cover the whole investigation period either. The images used were taken with either sensor type PS2 or PS2.SD [22]. The sensor PS2 consists of a four-band frame imager (red: 590–670 nm, green: 500–590 nm, blue: 455–515 nm, NIR: 780–860 nm) with a split-frame for the visible RGB (red/green/blue) and the NIR. PS2.SD consisted of a four-band frame imager (red: 650–682 nm, green: 547–585 nm, blue: 464–517 nm, NIR: 846–888 nm) with butcher-block filter providing separate stripes for blue, green, red and NIR. The GSD of 3.9 m leads to a higher resolution (3 × 3 m). Both RapidEye and PlanetScope provided atmospherically corrected imagery.

To calculate the Normalized Difference Vegetation Index (NDVI) [23,24], the Red (R) and Near Infrared (NIR) bands were used:

$$\mathrm{NDVI}=\left(\mathrm{NIR}-\mathrm{R}\right)/\left(\mathrm{NIR}+\mathrm{R}\right)$$

(1)

Subsequently, data were limited to the scene with the highest NDVI values in each month (maximum value approach) assuming that these images were least affected by cloud and haze. Image selection was undertaken with a preference for scenes captured in the middle of the month and around the same day each month to allow a constant observation frequency and to increase the comparability across years. However, this was not always feasible due to the limited availability of cloud-free satellite images. The 14 dates used are listed in

Table 2. Overview of the used satellite images with the satellite identifier and the sensor type. No usable images were available for May 2019.

Date	Satellite Constellation
15 April 2015	RapidEye
24 May 2015	RapidEye
12 June 2015	RapidEye
23 April 2016	RapidEye
11 May 2016	RapidEye
08 June 2016	RapidEye
22 April 2017	RapidEye
19 May 2017	RapidEye
11 June 2017	RapidEye
23 April 2019	RapidEye
05 June 2019	RapidEye
21 April 2020	PlanetScope
21 May 2020	PlanetScope
15 June 2020	PlanetScope

.

Due to the described inherent different spatial and spectral resolutions between RapidEye and PlanetScope imagery, and to reduce sources of bias in the NDVI data, PlanetScope data were only used for statistical data analysis when NDVI values were normalized and divided into quartiles. By using normalized NDVI values and a quantile-based approach, we evaluated general trends and patterns and accounted for the uncertainty in NDVI values due to interannual variations in crop phenological development and crop-specific optical properties.

For all statistical analyses and graphs, the statistical software R (version 3.6.3, Windows 10 [25]) was used. Masks were applied to the satellite scenes to select corresponding grid cells for each altitude class (Figure 1) and to subsequently extract information on soil classes (Figure 1) within the altitude classes (package: “raster” [26]). NDVI values were not normally distributed, which was tested with the Shapiro–Wilk test. Therefore, the non-parametric Kruskal–Wallis test was used to test for significant differences in the NDVI values between classes for each scene. Here, an analysis of variance was used to test whether independent samples belong to the same population [27]. As a post-hoc test, the non-parametric Wilcoxon signed-rank test was applied whenever significant differences were found to specify which groups significantly differed by testing two paired samples for their central tendencies [28]. To avoid type 1 errors in the Wilcoxon test, the p-value was adjusted with the Bonferroni method [29].

To compare NDVI values across years, the NDVI values were normalized by dividing all the NDVI values of each grid cell with the mean of all grid cells. Subsequently, the relative NDVI values of each scene could be classified into four different classes. To do so, the values were split into quartiles, and each quartile was assigned its own value (1 to 4). This processing resulted in images showing the classified relative NDVI values, ranging from low (class Q1) to high (class Q4) NDVI values.

First, the non-parametric Mann–Kendall trend test [30] was applied to test for the presence of a monotonic trend in the time series over time (R-package: “Kendall” [31]). Tests were applied for each month separately. This test is commonly used in environmental research [32,33]. Subsequently, a correlation matrix was created by using the “corrplot” package [34]. In the first step, 10% of the pixels per scene were subsampled and pairwise correlations between those pixels were calculated across the years, using the non-parametric Spearman rank correlation using the global “cor” function implemented in R. To quantify the changes in the spatial distribution and size of NDVI clusters between years, cross-tabulations (crosstabs) for all possible combinations of months across all years were computed.

3. Results

3.1. Climate

During the vegetation period, the total precipitation was between 99 and 117 mm. Only in 2017 (220 mm) was the precipitation sum comparatively high and significantly higher than the long-term average value (149 mm). However, this mainly resulted from higher precipitation during the late growing season in June and July, thus, after the observed period used for our study (Figure 2). In general, the climatic water balances (CWB) were rather positive in autumn (October/November) and winter until February/March, followed by negative CWBs during spring and summer (Figure 2). The observation years did not include the main Central European drought year, 2018 [35]. However, spring 2020 was drier compared with the other years in March (CWB = −21 mm). Further, the summed CWB for the months of April to June revealed the driest conditions in 2019 (CWB = −232 mm), followed by 2020 (−205 mm), 2016 (−196 mm), 2015 (−183 mm) and rather wet conditions in 2017 (−57 mm).

3.2. Variation of the NDVI in Altitude and Soil Classes

The total number of grid cells (n) of a single RapidEye scene was 2181. The total number of grid cells per category (soil, altitude) and the resulting area are shown in

Table 3. Size [m²] and grid cell number of each soil (S1–S3) and altitude (A1–A3) class with the definition of each class.

Class	Definition	Numbers of Grid Cell [n]	Area [m2]
S1	Loamy layer at <60 cm depth	951	23.775
S2	Loamy layer at 60–100 cm depth	351	8.775
S3	No loamy layer	879	21.975
A1	<53.5 m a.s.l.	752	18.800
A2	53.5–55.37 m a.s.l.	804	20.100
A3	>55.37 m a.s.l.	625	15.625

. Across all years, the lowest and highest NDVI values were recorded for altitude class A1 in June 2019 (0.47, SE: 0.002) and for A3 in May 2015 (0.88, SE: 0.002), respectively. With respect to the different soil classes, the distribution of the highest (S2, April 2017: 0.87, SE < 0.001) and lowest NDVI values (S3, June 2019: 0.49, SE: 0.002) was slightly less pronounced. In addition, as shown in

Table 4. Cross-table to show the overlapping grid cells of the altitude (A1–A3) and soil (S1–S3) classes. In the columns: S1–S3, in the rows: A1–A3. A1: lowest altitude, A3: highest altitude. S1: highest water holding capacity, S3: lowest water holding capacity.

Class	S1	S2	S3
A1	97	100	554
A2	461	171	174
A3	393	80	151

, A1 overlapped most frequently with S3 (554 of 751 same grid cells), A2 with S1 (461 of 806same grid cells) and A3 with S1 (393 of 624 same grid cells).

Results from the Kruskal–Wallis test indicated statistically significant differences (p < 0.05) of NDVI values between soil classes and altitude classes. To test for the source of these differences and to identify non-significant class pairs, the Wilcoxon signed-rank test was performed for all scenes. Out of 66 possible pairs of combinations, only 12 were not highly significant (p ≥ 0.05,

Table 5. All combinations of not highly significant (p-value ≥ 0.001) differences within the soil and altitude classes calculated with the Wilcoxon signed-rank test.

Year	Month	Combination	p-Value
2015	April	S2–S3	≥0.05
		A2–A3	1
	May	S1–S2	≥0.05
		A2–A3	≥0.05
2016	June	S1–S2	1
2017	April	S1–S3	1
		A1–A3	≥0.05
	May	A2–A3	1
	June	S1–S2	1
		A2–A3	1
2019	April	S1–S2	≥0.05
		A1–A2	≥0.05

). Four of the six non-significant soil class combinations were found between S1 and S2 and four of six of the altitude classes between A2 and A3.

3.3. Spatial Distribution of NDVI over Time

Based on the quartile classification of NDVI data, results showed that the northernmost area of the field had consistently lower NDVI values (most pixels belong to the first quartile Q1 in each scene) (Figure 3 and Figure 4) as compared with the rest of the field. In April and May 2017, the spatial clustering of NDVI values was much lower compared to all other years (Figure 3). A figure with all NDVI quartile maps can be found in the supplementary material (Supplementary Material Figure S1).

The Mann–Kendall trend test ruled out the presence of temporal trends in the NDVI distribution. Comparing monthly data across years using correlation analyses, it was noticeable that, overall, data from April had the lowest correlation values (mean: 0.26, SE: 0.039, n: 10), indicating the highest interannual differences in spatial patterns during early spring. The highest interannual agreement was found for June recordings (mean: 0.64, SE: 0.025, n: 10). The cross-tabulation matrix (

Table 6. Cross table considering all months (April–June). (A): All years, (B): Only years with triticale planted (2015/2016/2020). Each crosstab shows the percentage of grid cells per defined class (1–4) in the columns for one, and in the rows for another, scenery. One hundred percent correspond to all grid cells per row (i.e., quartiles). With this approach, it was possible to quantify the number of spatially equally distributed grid cell classes between two sceneries as well as the change in classes. The results of the cross table showed that the category with the lowest NDVI values (Q1) remained most frequently the same over the years, followed by the class with the highest NDVI value. Consistency in the NDVI category was lowest in April (max. 45%) and highest in June (up to 77%).

(A)
Quartile	1 [%]	2 [%]	3 [%]	4 [%]
1 to	60.1	19.3	11.7	8.8
2 to	19	33.2	27.1	20.7
3 to	11.9	27.1	32.8	28.2
4 to	9.1	20.3	27.8	42.7
(B)
Quartile	1 [%]	2 [%]	3 [%]	4 [%]
1 to	70.2	18.4	6.8	4.6
2 to	16.6	39.7	26.7	17
3 to	7.7	27.3	36.6	28.4
4 to	5.7	14.7	29.1	50.4

) showed that the category with the lowest NDVI values (Q1) had the highest spatial consistency across years (60–70%), followed by the class with the highest NDVI value (Q4: 42–49%). Consistency in the assignment to the different NDVI quartiles was lowest in April (max. 45%) and highest in June (up to 77%).

4. Discussion

4.1. Spatial Distribution of NDVI

Since the differences in average NDVI values between soil classes as well as between altitude classes were statistically significant in 60 out of 66 possible pairwise comparisons, each soil type and altitude class could be assigned to an average high or low NDVI value for each scene. The fact that both soil type and elevation affected crop growth is consistent with the literature [4,12]. Due to the large spatial overlap of classes, final statements about the ranking of the influence of elevation or soil on NDVI values cannot be made. A study that has been carried out in the same field evaluated the influence of altitude and soil on plant growth as well as harvest yield in 2020 and showed that both factors have a major impact on plant growth depending on the crop growth stage [16]. A combination of soil type S3 (no loamy layer in the subsoil, thus, lower available water capacity) and elevation class A1 (downslope) appears to represent the worst conditions for plant growth and grain yield at harvest [16]. This combination is true for most of the grid cells in the experimental field. These findings were confirmed by our study in terms of low NDVI values in the late growing season, which has been associated with lower grain yields [36,37,38,39]. Results on altitude effects contrast with those from studies where low-lying field sections benefit from their topographic position in terms of water availability [39]. This highlights the need for knowledge of vertical soil profile properties for understanding topography-induced effects on soil moisture and NDVI patterns [40].

A lower influence of the different soil and elevation classes on the NDVI values of rye and thus a spatially more heterogeneous distribution of rye NDVI values in 2017 compared to the other years might be attributed to both, (1) a higher summer water balance compared with the other years and (2) the different response of winter cereals to environmental growing conditions [41]. Winter rye, grown in 2017, is well known as a less demanding crop in terms of environmental conditions as compared with wheat; thus, NDVI patterns are likely less clustered.

In summary, results on the temporally persistent spatial distribution of NDVI classes suggest the presence of generally high- and low-yielding areas. In consequence, a spatially targeted crop management, e.g., with respect to planting density or fertilizer and pesticide application, will most likely allow for more resource-efficient land management at our site.

In general, the spatial scale of Rapid-Eye satellite data was suitable for obtaining information on this area. Using UAV data, future studies might test for an optimum spatial scale to assess the within-field heterogeneity for guiding specific management strategies [39] and help in reducing uncertainty in interannual comparisons caused by species-specific phenological development. This study used data from different winter cereals grown within the same fields within a crop rotation. Plants might have differed in terms of their physiological development and greenness in each year [42]. While this was accounted for by using normalized data and a quantile-based approach, uncertainty resulting from species-specific plant and canopy characteristics cannot be completely ruled out.

4.2. Influence of Weather Conditions on Spatial Patterns of NDVI at Field Scale

Due to the limited number of observation years with rather low differences in the climatic water balance (Figure 2), conclusive statements on the influence of different weather conditions (in particular, drought stress) on the spatial distribution of NDVI values have to be considered with care. Even the driest year (2019, April to June: CWB = −232 mm) did not lead to pronounced differences in the spatial distribution of NDVI values compared to the other years. The year with the highest CWB (2017, −57 mm) showed a more heterogeneous distribution (lowest clustering of NDVI values) (thereby rather confirming). However, in 2017, the field was cultivated with winter rye, which is known for being a robust and low-demanding crop. In consequence, the lower spatial clustering of NDVI values in 2017 could be either due to more humid conditions or due to the higher adaptability of winter rye to changes in soil conditions.

Under Mediterranean climate conditions and generally in arid or semi-arid climates, spatial dynamics in NDVI values have been related to differences in the soil water capacity and to drought indices [43,44,45]. Contrastingly, at our site, under mild temperate conditions with optimum fertilizer supply and using varieties optimized for local conditions, the influence of weather conditions might not be assessed at the sub-field scale for annual cereal crops. Satellite NDVI values are affected by both biomass and greenness. These factors might have differed between crops and varieties each year, thereby adding uncertainty to the results. As an example, in spring, when NDVI values are not yet saturated, nitrogen limitations have been found to be critical for canopy greenness, whereas the importance of soil water availability increases in later summer months. Such observations illustrate the complexity and nonlinearity of environmental effects on NDVI patterns and highlight the importance of spatially detailed studies at the sub-field scale to understand environmental effects on vegetation dynamics as observed from space [39].

5. Conclusions

For a field in North-Eastern Germany, spatial patterns of NDVI values from winter cereals are affected by the spatial distribution of soil available water capacity, as derived from the depth to a loamy layer, and topographic position. The quartile distribution of normalized NDVI values illustrated temporally stable spatial patterns across years, especially in the areas with poor growing conditions, which can be linked to high- and low-yielding zones. Our study highlights the strong influence of the spatial variability of soil and terrain characteristics, especially for those field areas that have poor growth conditions (consistently low NDVI values). Differences in the climatic water balance can contribute to explaining interannual differences in spatial patterns, i.e., clustering of values (with lower clustering in wetter years); however, species-specific differences in the sensitivity to water stress have to be considered. Satellite data provided by RapidEye or PlanetScope with a high spatial resolution are suitable to detect within-field differences in agricultural growing conditions for supporting optimized variable rate management.